Bayesian Prestack Inversion Based On Multipoint Geostatistics

At present,in the process of maximizing the potential of old oil and gas fields and efficiently developing new oil and gas reservoirs,there is an important challenge,that is,how to make full use of core,logging,seismic,geological and other data information of different scales to realize the high-precision reservoir characterization of complex structural reservoirs.In this paper,taking advantage of multipoint geostatistics to describe the correlation between multiple points in space,we first integrate the log data and known geological information to model reservoir lithofacies.We use the widely used SNESIM algorithm,which uses search tree to store the dynamic data greatly to speed up the simulation speed,the simulation results and real data have a high degree of coincidence.Then we study the Bayesian prestack inversion method and use seismic data to retrieve elastic parameters,such as velocity of longitudinal and shear waves and density of the target geological body.In order to solve the instability and multi-solution problems of seismic inversion,as well as the limitation of seismic data frequency band and the problems of noise,regularization constraints should be added to the inversion process.In this paper,combining the advantages of Tikhonov method and total variational method,we use the Markov random field constraint.By using the equivalence between Markov random field and Gibbs distribution,we use the Huber function as the potential function,and reasonably set the threshold and regularization parameters.It can not only effectively suppress noise to smooth the layer,but also can protect the edge information.Finally,we propose a Bayesian prestack inversion method based on multi-point geostatistics by combining multi-point modeling with prestack inversion by using phase-controlled Sequential Gaussian Simulation.This method can combine multi-scale information,such as seismic data,logging data and geological information,to accurately characterize the reservoir.The simulation results of three elastic parameters are obtained by phase-controlled Sequential Gaussian Simulation with multipoint modeling lithofacies and known well data.A slight smoothing of the simulation results is used as the initial model for inversion.The results of three parameter inversion are stable and continuous,and accord with the real data very well.The effectiveness of the above method is verified by model tests.