Geostatistics Inversion Method Research Under The Constraint Of Seismic Data

The basic task of geophysical exploration is to obtain the relevant physical parameters and elastic parameters information of underground medium through the study of the observation seismic data,then geophysicists can interpret the capability of reservoir and give an evaluation.As an important way to accomplish this task,geophysical inversion has important practical significance in reservoir characterization.Conventional deterministic inversion methods mainly depend on the seismic data as constraints,which is difficult to depict thin interbed reservoir spatial distribution.Stochastic seismic inversion using welldata as the constraint.By combining geological statistical modeling and seismic inversion,it can break through the seismic data bandwidth limitations,and effectively reduce the uncertainty of inversion problem.As a result,stochastic inversion can largely improve the vertical resolution of inversion results.In this paper,we studied the geostatistical inversion method under the constraint of seismic data,by the effective combination of geostatistics,seismic inversion theory and the optimization algorithm,this method can be used to solve nonlinear geophysical inverse problems.This algorithm is constrained both by seismic data and logging data for geological statistics.Then under the bayesian framework,we can statistically analyse the posterior probability distribution and obtain model parameters directly.Then we adopt Markov Chain Monte Carlo method for global optimization.As a result,geophysical parameters of underground reservoir can be obtained.On the basic theory of geostatistics stochastic modeling and seismic inversion,we lead to the stochastic seismic inversion.Under the bayesian theory framework,we designed a process of stochastic inversion method,and take several model tests for this method,the inversion results proved the effectiveness of this method.At last,we extended it to prestack multiwave joint inversion making full use of the p wave and ps wave information.Model tests prove the feasibility of this method.