Carbonate reservoirs are an important aspect for improving the revealed reserves and productions of oil/gas resources in China.However,such sort of reservoirs usually develops complex pore structures,which may influence the accuracy of seismic prediction for petrophysical parameters.This work proposes a seismic rock-physics simultaneous inversion method for petrophysical and pore parameters with the Gaussian mixture model.Based on the differential effective medium model and the Gassmann equation,a linearized forward operator is derived to quantitatively relate porosity,fluid saturation,and pore aspect ratio to the elastic parameters.To characterize the statistical variations within lithofacies,the Gaussian mixture model is introduced to describe the joint prior probability distribution of petrophysical and pore parameters.The analytical expression for the posterior distribution of the objective parameters is obtained with the linearized model based on the Bayesian inverse theory.Pore aspect ratio is inverted at a well location so as to provide reliable prior constraints of the pore parameter,and an iterative Bayesian inversion algorithm is adopted to improve the forward modeling accuracy.The proposed method treats pore aspect ratio as an objective parameter to account for the spatial variations of pore structures in carbonates,and employs the Bayesian linear inversion to improve the accuracy and efficiency of seismic prediction for petrophysical parameters.Numerical tests indicate that the accuracy of inverted petrophysical parameters by the proposed method is significantly improved,and the method is less affected by the initial models and elastic parameters to some extent,compared to those by the conventional method.The application to the 3D data validates the method,wherein the porosity result well indicates the spatial distribution of potential reservoirs.
关键词
地震岩石物理反演/孔隙纵横比/碳酸盐岩储层/贝叶斯线性反演/高斯混合模型
Key words
Seismic rock-physics inversion/Pore aspect ratio/Carbonate reservoir/Bayesian linear inversion/Gaussian mixture model
维普
万方数据