Seismic Inversion Based On Geostatistics

Geostatistics plays an important role in reservoir characterization and modeling.Geostatistical stochastic simulation can give multiple equal probability modeling results for uncertainty analysis.However,in practice,the horizontal resolution is not high due to the limited number of wells,and the cross-well simulation results usually have large uncertainties.The geostatistical random inversion of seismic records as a constraint can well reflect the changes in the large areas between wells.Compared with the random simulation,the reliability of modeling results is improved,and the uncertainty is reduced.The geostatistical random inversion matches the synthetic seismic record of the simulation result with the actual seismic record to obtain a modeling result that is consistent with the seismic record.Improve the accuracy of reservoir characterization.Based on the basic theory of Geostatistics,the dissertation develops a random geostatistical simulation and a randomized inversion method for Geostatistics.Not only analyzes and tests the factors that affect the random simulation results of Geostatistics,but also clearly defines the role and physical meaning of different parameters in stochastic simulation.It also studies the effects of different well density on the simulation results.The number of well data is of great significance for improving the accuracy of reservoir modeling results.In addition,a stochastic inversion method for Geostatistics was studied.Based on the sequential Gaussian simulation,the idea of Quantum Annealing was introduced.After each random simulation,a Quantum Annealing inversion was performed to match the seismic records.As a new condition data,it is added to the condition data group to simulate the increase of the number of channels,and more and more conditional data are available.The ability to simulate and invert constraints in the next track is enhanced,reliability is improved.Compared with the more commonly used Simulated Annealing algorithm,the Quantum Annealing algorithm has better convergence performance,and it is not easy to fall into a local minimum and can obtain better inversion results.The comparison of these two methods fully demonstrates the effectiveness and validity of the proposed method,and obtains better inversion results,providing effective reference information for reservoir characterization and fine description.