| In recent years,the field of oil and gas exploration has been deepening,the exploration methods are becoming more and more innovative,the exploration section is shifting from shallow to deep,the geometry of oil and gas reservoirs is shifting from simple to complex structures,the object of oil and gas exploration is gradually starting to change from tectonic to lithological reservoirs,the conventional oil and gas resources are continuously depleted,and the shale oil and gas resources are getting more and more attention.However,unconventional reservoirs are characterized by complex pore-permeability conditions,strong reservoir inhomogeneity and complex dispersion attenuation mechanisms compared with conventional reservoirs,so conventional petrophysical theoretical models and methods are difficult to obtain accurate results in cross-wave prediction and attenuation estimation,and need to be improved in a targeted manner.This thesis summarizes the boundary model for rock matrix,the inclusion model based on the equivalent medium theory,the Gassmann theory for fluid and solid replacement and its various extensions,and the macroscopic flow,microscopic flow,and mesoscopic flow models developed from the porous medium theory,through studying the theory and methods of seismic rock physics modeling.theoretical basis.Starting from the "three-step" Xu-White rock physics combination model,combining the existing experimental results,analyzing the shortcomings of fixed pore aspect ratio,considering the non-uniformity of microscopic pores in the longitudinal direction and the continuity in thin layers,and combining with the whale optimization algorithm,the adaptive Xu-White model is proposed.The SCA model with a single pore structure is compared with the conventional Xu-White model to verify the accuracy of the model transverse wave prediction.On this basis,the high dispersion and strong attenuation phenomena of shale reservoir seismic frequency band were taken into consideration,and the effects of pore permeability,viscosity,saturation and other parameters on the dispersion attenuation of macroscale,microscale and mesoscale models were studied by numerical simulation.The model was applied to the shale reservoir in the study area,and the quantitative characterization of the seismic frequency band attenuation was completed while ensuring the accuracy of longitudinal and transverse wave velocity prediction by combining the measured data at different scales.At the same time,combining with viscoelastic theory,a rock physical template is established to link the attenuation parameters and reservoir characteristics,providing theoretical guidance for reservoir prediction and fluid identification.This thesis combines the rock physical boundary model and the data-driven model in an a posteriori constrained manner to carry out interdisciplinary cross-application,and proposes a "data + mechanism" dual-driven fusion model,which makes full use of the advantages of Cat Boost,Light GBM and other machine learning methods in the nonlinear fitting of multivariate relationships,and overcomes the risk of overfitting caused by small sample data,effectively improving the prediction accuracy of the model for elastic parameters and the overall generalization ability of the model,and providing a solid data basis for guiding pre-stack inversion and reservoir prediction with high accuracy. |
