Study On Geophysical Joint Inversion Method Locating Deep Thermal Storage In Gonghe Basin,Qinghai Province

Gravity and magnetic exploration are the method of reconstructing subsurface density and magnetization images by gravity and geomagnetic field data.It is characterized by low cost and high efficiency compared with other geophysical methods,such as seismic and electromagnetic.Therefore,gravity and magnetic data are generally two-dimensional data,while electromagnetic and seismic data are often profile.However,due to the ill distribution of sensitivity matrix and the immaturity of the inversion algorithm,gravity and magnetic inversion is more regarded as the methods of low depth resolution.In order to improve the reliability of potential field data inversion,it is now the trend to combine with other information,such as geological interpretation,other geophysical data,borehole information,into the inversion process of potential field data.The key point of improving resolution of 3D images reconstructed by gravity and magnetic data is the joint of electromagnetic or seismic information into the inversion.This paper first studies basic theory of forward and regularization inversion method of potential field data,including the basic Tikhonov inversion and weighted inversion.After known the reasons for the stability of inversion in weighted space,the method of constrained inversion in weighted space is proposed.Model validations show the new proposed method demonstrates more flexible performance.This paper also optimizes the flow of the conjugate gradient algorithm,avoiding repeated computations and use of large matrices,which greatly improving computational efficiency.Then this paper studies performance of cross entropy item in potential field data inversion.The differential calculation method for cross entropy vectors and scalars is implemented in two group models to study the role played by cross entropy vector in the inversion.A method for calculating gradient of cross entropy based on backward propagation is proposed and its validity is verified by iterative experiments.The validity of the cross entropy joint inversion is verified through a 3D model experiment with fixed model constraints.This paper proposed subdimension inversion method to joint multiple information.Instead of solving the equation directly in the original space,subdimension method first projects the equation to a lower dimensional space and solves the equation in the lower latitude space.Multiple solutions are obtained through different projections,and finally these solutions are integrated to obtain the final solution.Model experiments show that the subspace inversion algorithm solves the depth discrimination problem of field data inversion while enabling a new form of joint inversion by introducing external information into the inversion by space divisions.Finally,this paper studies the distribution of geothermal reservoirs and the location of heat sources in Gonghe region using the MT inversion results as constraints.Cross entropy and subdimension methods are both used,which is significant for the further development of geothermal in the region.