Grey Wolf Optimization algorithm with random local optimal regulation and first-element dominance

Abstract

Due to the classical Grey Wolf algorithm GWO does not consider the characteristics of the local information of individual in population, a novel local random optimization strategy is proposed to make up for the defect of GWO. In this method, several points in the neighborhood of the current location of each individual are selected at random in the axial direction as candidates, and the best points are selected to participate in the renewal decision of the individual. Furthermore, in our experiments, a special first-element dominance characteristic is found and can greatly improve the combination effect of global and local information. In order to ensure that all constraints are not violated in the process of constraint optimization in industrial design, the random mixed population initialization method is proposed to generate population individuals that meet the constraint requirements and contain boundary values randomly. In addition, a treatment method of shrinking in a specific direction is proposed for dealing with individuals who cross the boundary. Experimental results on several test function sets show that compared with recent improved algorithms for GWO, the proposed algorithm has obvious advantages in fitness value, convergence speed and stability.