The Research Of Multiscale Seismic Inversion

As the deepening of exploration degree,subtle hydrocarbon reservoirs and fluvial facies thin reservoirs have become key exploration areas.These areas are characterized by complex geological background,thinner reservoir thickness and small horizontal distribution changes.Seismic data have low signal-to-noise ratio(SNR)and narrow effective bandwidth.In addition,the inversion method itself also has ill-posedness,seriously affected by local minimum value,so as to reduce the resolution and reliability of wave impedance inversion.Conventional seismic inversion is difficult to identify thinner reservoirs,which needs to improve the quality of seismic data and explore a new inversion method.The reserving space is affected by sedimentary environment with multi-scale characteristics,and seismic data mainly reflect the reflection characteristics of large-scale sequence-body thick layer.Influenced by thin-layer tuning effect,reflection energy of small-scale sequence-body thin layer is suppressed to reduce the seismic resolution.Wavelet transform has good time-frequency concentration able to reflect the subtle multi-scale information of seismic data clearly in time-frequency domain.In this paper,three-parameter wavelet is taken as a kernel function for time-frequency analysis on seismic data to decompose seismic data into large,medium and small-scale data volumes,weaken the tuning function effectively,enhance the relative high frequency and relative low frequency of seismic data and improve the resolution.In addition,the diffusion filtering technology is used to efficiently attenuate random noises of 3D seismic data,protect the stratigraphic boundary information,improve the quality of seismic data and provide the reliable data for subsequent inversion.Based on Bayesian Theory and multi-scale corresponding relationship between seismic data and sequence body,a stable and efficient multi-scale inversion method is adopted.Target function scale is continuously shrunk to approach the global minimum point,so as to realize the whole inversion.The whole process is carried out by level.The large-scale seismic inversion result is taken as the model constraint of meso-scale inversion,and meso-scale inversion result is the model constraint of small scale inversion,so as to complete the whole inversion finally.Based on the research on classified inversion in this paper,the inversion process is written in matrix form to incorporate three steps of inversion for one step,greatly improving the operation efficiency.For elastic impedance inversion,angle stacked data volumes are used to integrate the fidelity of AVO inversion and the stability of post-stack inversion.The result can be applied to extract multiple elastic parameters for reservoir prediction and hydrocarbon identification.After desired effect is achieved for multi-scale post-stack inversion,this method is promoted to pre-stack elastic impedance inversion.Model tests on theoretical and actual data show that the multi-scale inversion has a step-by-step optimization property.The resolution of result is higher than conventional inversion,validating the advancement and practicability of the method.The inversion results of practical application work area indicate that compared with conventional sparse pulse inversion method,the resolution of multi-scale inversion result has substantially improved to identify thinner effective reservoirs and better serve hydrocarbon exploration,development and production.