Prestack Seismic Probabilistic Inversion Driven By Facies

With the development of oil-gas exploration in China,the types of hydrocarbon reservoirs are becoming more complex,especially in recent years,the exploration and development are advancing to the fields of subtle lithologic reservoirs,thin interbedded reservoirs and unconventional hydrocarbon reservoirs.As the core and frontier subject of reservoir geophysics,prestack seismic inversion is the most critical approach to the quantitative characterization of complex reservoirs,the prediction of physical properties,and the hydrocarbon identification.Our paper focuses on the prestack seismic probabilistic inversion in complex reservoir prediction and fluid identification,and mainly carries out three parts: 1)probabilistic seismic inversion driven by lithofacies,aiming at realizing the synergistic identification of "model parameters" and "discrete lithofacies";2)fluid-facies driven prestack seismic probabilistic inversion,aiming at realizing the synergistic prediction of "elastic parameters" and "discrete fluid-facies";3)prestack seismic probabilistic inversion driven by rock physics to realize the simultaneous prediction of reservoir physical parameters,rock matrix,dry rock skeleton and discrete lithofacies.The innovations are as follows,Seismic inversion is the most important quantitative evaluation method for acquiring model parameters and lithologic information of subsurface media.Conventional seismic inversion often predicts the "model parameters" and "lithology" independently,ignoring the influence of subsurface lithology on model parameters.The errors of prior knowledge introduced will seriously affect the accuracy of seismic inversion and lithology prediction.In this paper,considering that the prior probability distributions(PDF)of model parameters are affected by subsurface lithofacies,the mixture prior PDF is utilized.Under Bayesian framework,the explicit solution of posterior PDF is derived,which is constrained by four conditional data sets: time-domain,frequency-domain seismic data,low-frequency prior and the known model points.The non-linear boundary constraints are applied in the estimation of elastic parameters.Sequential simulation and stochastic sampling are utilized to realize the simultaneous prediction of "model parameters" and "lithofacies".The introduction of low-frequency prior and boundary constraints improve the stability of seismic inversion.Besides,sequential simulation and stochastic sampling improve the inversion resolution.Seismic fluid identification is the important contents of reservoir geophysics.Under the guidance of seismic rock physics and pre-stack seismic inversion,it is utilized to describe the fluid-bearing characteristics of complex reservoirs.Based on the mixture prior PDF of elastic parameters,we derive the standard posterior PDF represented by one high-dimension mixture probability model.Then,one novel prestack seismic probabilistic inversion based on DE-MCMC model is proposed.The combination of differential evolution and MCMC model has the characteristics of global optimization,the probability estimation,and uncertainty analysis.Multiple MC chains run parallel and obtain the multiple solutions of model parameters,which realize simultaneous identification of elastic parameters and fluid-facies.Seismic rock physics play a bridge role between elastic modulus,reservoir physical properties and fluid parameters.An approximation of Xu-White rock physics model is constructed,and the influence of rock-skeleton and pore-fluid on rock modulus is investigated.Besides,it overcomes the limitations of linear fitting between elastic parameters and physical parameters.Then,one AVO approximation characterized by Jacobian and Hessian matrix is deduced,and the contributions of porosity,pore fluid and physical parameters to AVO reflectivity are analyzed.These analyses illustrate the feasibility of the direct estimation of physical parameters.A facies-driven seismic probabilistic inversion based on DE-MCMC is proposed,which realizes the simultaneous estimation of rock physical parameters,elastic parameters,dry rock-skeleton modulus and discrete lithofacies.These proposed approach is applied to the broadband seismic collected from east of China.Combining with broadband seismic cascade inversion in complex frequency domain,the potential relationships between low frequency energy in complex frequency domain and Laplace attenuation coefficients are analyzed.The penalty function in complex frequency domain is constructed.Two inversion techniques are created under Bayesian inference:(1)Broadband seismic inversion based on maximum a posteriori estimation;(2)prestack probabilistic inversion driven by rock physics.Several examples illustrate that the proposed approaches have strong high-low frequencies recovery ability,improve the inversion resolution,lay a theoretical foundations for broadband seismic inversion,facies identification,quantitative precition of physical parameters,which have broad application prospects.