Parameter fitting of variogram based on hybrid algorithm of particle swarm and artificial fish swarm

Abstract

Variation function is an important tool for describing the spatial correlation characteristics of regionalized variables in geostatistical methods. Variation function modeling is an important part of kriging interpolation and will directly affect the accuracy of the final interpolation result. The purpose of this work is to address the shortcomings of traditional variogram fitting methods, introduce particle swarm algorithm and artificial fish swarm algorithm under swarm intelligence framework, and design a variogram parameter fitting based on the hybrid algorithm of particle swarm and artificial fish swarm method. With this method, the minimum difference between the variation function fitting model and the given experimental variation value is utilized as the optimization goal. An appropriate objective function is set to convert it into a minimum problem. The hybrid algorithm has a strong search ability and convergence, as well as the ability to obtain the satisfactory fitness values. By comparing the results of the VARFIT fitting and the results of the optimization algorithm, it can be concluded that the absolute deviation of the fitting results of the optimization algorithm is 3.39 lower than the results of the VARFIT fitting. Compared with the traditional variogram modeling approach, this method has a strong optimization ability and high precision, and can effectively realize the automatic fitting of variogram parameters.