Prestack Seismic Inversion Of Deep Reservoirs

With the deepening of exploration,the proportion of proven shallow reservoirs has been increasing for years.Commercially valuable oil and gas reservoirs have been decreasing.So the target of exploration is turning to the reservoirs at intermediate and deep depth.Prediction of deep reservoirs has become the focus of research on geophysical exploration.Inversions based on seismic amplitude information are commonly used.Reliable inversion results contribute to the increase of drilling success ratio and the reduction of exploration costs,which is very important for the reservoir exploration and evaluation.Nowadays,there are few seismic inversion methods for deep reservoirs.And applications of traditional methods in deep reservoir exploration are not good enough.This thesis proposes prestack seismic inversion methods according to the characteristics of deep reservoirs.The offset(incident angle)ranges of deep reservoirs are limited.Traditional prestack seismic inversion methods always need data in large offsets(incident angles).So it's not good to use traditional methods in deep reservoir inversions.Similar to the derivation of Smith-Gidlow two-term approximation,this thesis starts from the approximations of Zoeppritz equations and derived an approximation for seismic data in small offsets(incident angles).Based on this,with the help of relationships of elastic parameters,new approximations for seismic data in small offsets(incident angles)with different elastic parameters are also derived.Based on the Bayesian theory,an elastic impedance inversion method for seismic data in small offsets(incident angles)is established.The model test and real data example proved that accurate results can be inverted without seismic data in large offsets(incident angles).In the process of prestack seismic inversions,the error of angles and the problem caused by the error have long been ignored.In the approximations used in prestack AVO/AVA inversions,the angle is the average of incident angle and refracted angle.Angle domain common imaging point gathers,which are the seismic data used in prestack inversions are obtained according to the incident angles.The difference between average angle and incident angle causes errors.The seismic data of deep reservoirs,with low signal-to-noise ratio and low preservation,causes greater errors to inversion results.This thesis starts from an approximation of Zoeppritz equations.An angle-optimized approximation with incident angle is developed to solve the inversion error caused by angle error.Based on this approximation,a corresponding prestack AVO inversion method is built.Accompanied with the newly-built inversion method for small angle seismic data,an angle-optimized probabilistic prestack inversion method for small angle seismic data is established.Both the model test and real data example show higher accuracy of inversion results.This lays foundations of data for reservoir prediction and fliud discrimination.The effect of compaction,pressure and temperature to the deep reservoir leads to different characteristics from the shallow ones,which causes the bad performance of traditional fluid discrimination methods.According to the porous elastic media theory,fluid sensivity of traditional fluid indicators are tested.The fluid modulus is proposed to be used as fluid indicator of deep reservoirs.An approximation containing the fluid modulus is derived.With four-variable Cauchy distribution as the priori and four-variable Gaussian distribution as the likelihood function,a Bayesian inversion method is established.Both the model test and real data example proves that fluid modulus is a good fluid indicator for deep reservoir.