| Magnetotelluric sounding is an important method to study the electrical structure of the earth by using the natural electromagnetic field as the field source.Because of its high efficiency,low cost,convenient construction,and high resolution,magnetotelluric sounding has been applied to the exploration of various resources,such as geothermal,water resources,mineral resources,oil and gas,etc.At present,with the deep research of magnetotelluric theory,many new problems have appeared,so they need to be solved and improved,and these problems involve the research of forward numerical simulation and inversion algorithm.This paper mainly discusses the forward numerical simulation of one dimension,two dimension and the regularization inversion algorithm of hybrid intelligent algorithm.Forward modeling is the basis of inversion.Based on the Maxwell equations,this paper deduces the calculation formula of magnetotelluric forward modeling of homogeneous half space and layered medium,which lays a foundation for the later inversion calculation.At the same time,the boundary value problem and the variational problem of two-dimensional magnetotelluric are derived,and the rectangular mesh is used for subdivision,and the biquadratic interpolation finite element method is used to solve the corresponding boundary value problem.According to the multi frequency characteristic of forward modeling,the parallel computing of vectorization is used to improve the computing speed of forward modeling.In order to verify the correctness of the two-dimensional forward modeling,a typical geoelectric model is used in this paper.The results show that the numerical simulation of the magnetotelluric forward modeling is fast and accurate.The inverse problem of magnetotelluric sounding is ill posed.In order to solve this problem,this paper introduces the regularization method to constrain the objective function,so that the ill posed inverse problem has a stable inversion result.This paper combines particle swarm optimization,differential evolution and Nelder-Mead to design a nonlinear hybrid intelligent algorithm.Hybrid intelligent algorithm not only uses the ability of particle swarm optimization and differential evolution algorithm to seek global optimization,but also uses Nelder-Mead optimization algorithm to further develop local optimization results,which solves the problems of linear algorithm and single nonlinear intelligent algorithm easily falling into local minimum value and low efficiency of optimization.Combined with the typical multi-dimensional function,the efficiency of the hybrid intelligent algorithm is verified.Then,the framework of the hybrid intelligent algorithm is constructed,and the objective function is set according to the apparent resistivity,impedance phase and inversion characteristics.In order to obtain stable inversion results,the regularization term is added to realize the completely nonlinear inversion.Then the inversion algorithm of different geoelectric models is verified.The results show that the hybrid intelligent algorithm has obvious advantages over the single fully nonlinear algorithm and linear algorithm.Finally,the hybrid intelligent algorithm is applied to the inversion of a measured point data of Jilin Huapichang geothermal field,and the inversion results are compared with the inversion results of the traditional inversion algorithm,which shows that the inversion results of the algorithm in this paper are more consistent with the actual exploration results,which proves that the hybrid intelligent algorithm has a high accuracy in the inversion of magnetotelluric measured data. |
