Study Of Rockphysics And Inversion Methods For VTI Anistropy Shale Reservoir

Stratigraphic anisotropy in the crust is widespread,especially in areas where shales are developed.From inversion and migration to formation interpretation and reservoir characterization,anisotropic parameters play a significant role and are an important source of information.The following strategies have been developed and proposed to address the unique characteristics of anisotropic shale strata in terms of rockphysical modeling,parameter prediction and inversion.A rockphysical model is developed for anisotropic shale strata with diverse constituent mineral compositions,complex pore fracture structures,and inhomogeneous fluid distribution.The model can input a variety of mineral compositions,takes into account the structure of pore fractures,and also incorporates the capillary pressure coefficient to regulate the modulus of mixed fluids.The capillary pressure coefficient describes the distribution of subsurface fluids(uniform distribution or patchy distribution)and simulates the subsurface conditions more accurately.Using the established rockphysical model,S-wave velocity was predicted in the actual work area and the results were in good agreement with the actual logging curves.Before performing the anisotropic pre-stack inversion,an initial model is necessary to be established.The obtention of anisotropic parameters is a difficult problem.There are two common methods,one is the inversion based on the rockphysical model,and the other is to get them by logging method.Both of these methods have certain limitations.Considering the existence of relationship between logging curves,the anisotropy parameters can be calculated using the deep learning method.Based on this,a deep multitriangular kernel extreme learning machine optimized by the flower pollination algorithm is derived to calculate the anisotropy parameters.This algorithm incorporates multicore learning method,which can improve the accuracy especially in complex data condition;the parameters of the neural network are optimized by flower pollination algorithm,which reduces the influence of human subjective factors and facilitates the training of the network.After comparing with the unimproved algorithm,the accuracy and robustness of the improved method are verified.The algorithm was applied to the actual work place data,and a better result was achieved.The anisotropy parameters are used as calculation results for a variety of applications such as lithology identification,fracture prediction,and migration,in addition to calculating the initial model.The VTI anisotropic AVO inversion based on the exact reflection coefficient equation can obtain accurate information of subsurface elastic parameters(velocity,density,anisotropy parameters,etc.).At present,isotropic linear approximation equation,isotropic exact reflection coefficient equation,and anisotropic approximation equation are commonly used for pre-stack AVO inversion.In shale strata with strong anisotropy,these methods may have large errors.In order to obtain more accurate elastic parameters,the exact reflection coefficient equation for VTI media should be used for inversion.In addition,an improved Cauchy constraint based on Bayesian theory is added in the inversion objective function,which enhances the robustness and protecting the energy information of weak reflections and improving the resolution at the same time.After the inversion stability and ambiguity analysis,fast and stable solutions are performed by generalized linear inversion theory and iterative reweighted least square algorithm.The algorithm was applied to the synthetic data to check the noise immunity performance.And finally applied to the actual data,the inversion results of elastic parameters were obtained in good agreement with the logging data.The frequency-dependent AVO inversion can directly obtain the subsurface fluid distribution condition,but the factors that cause the frequency variation of subsurface velocity are not only the effect of fluid,but also cause velocity anomalies due to improper processing of seismic data and other reasons,which can produce many artifacts in the inversion results.For the VTI anisotropic shale strata,the frequency-dependent AVO inversion is extended to the anisotropic case.Firstly,after rockphysical analysis,the Pwave velocity dispersion attribute as the anisotropic frequency-dependent AVO inversion result can also indicate the velocity dispersion location,and the anisotropic dispersion properties are dispersive and related to anisotropy.Based on the idea of finding fluids in shales strata,the P-wave velocity dispersion attributes and anisotropic dispersion attributes are fused to locate the position where both dispersion and anisotropy are strong.The guiding idea of the fusion method is the higher-order Markov random fields semantic segmentation.Finally,the method is applied to the actual work area.The fused attribute is capable of locate the location of the subsurface gas-bearing region accurately.