2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Application of Bayesian Stochastic Inversion Based on Frequency Divisions Reconstruction in Reservoir Prediction


The conventional deterministic inversion method can recover the strong reflector information well, the inversion results have good lateral continuity. However, limited by the limited bandwidth of seismic data, conventional inversion method inevitably has the shortcomings of low resolution, so the thin interbed cannot be effectively characterized. Bayesian framework theory can effectively integrate different types (logging, seismic, geology), data signals of different scales interest, especially the introduction of model prior information, greatly improves the accuracy and reliability of the inversion. Due to the posterior probability distributions are generally complex high-dimensional distributions, and it is difficult to describe their properties directly. With the Bayesian theory and the geostatistics, this paper proposes a Bayesian stochastic inversion method based on Gr Quasi-Acoustic frequency divisions reconstruction. It consists of configuration modeling and stochastic simulation inversion. The configuration model is to extrapolate the medium and high frequency logging information through the medium frequency seismic information, and replace the variogram with the reflection configuration to describe the variable space structure to ensure reasonable interpolation of wells in the same phase band. On the basis of the configuration model, the stochastic simulation corrects the inversion results of the full frequency band by the phase-controlled random high-frequency disturbance and the constraints of the seismic synthesis record. In this study, the principle of bayesian inversion method is analyzed in detail and it is proposed that Bayesian inversion is implemented by using seismic data, then a model is established based on the characteristics of Bohai Q oilfield. Under the guidance of geology research and depositional development model, using the method of well logging constraint seismic inversion to compensate the high frequency and low frequency, and to complement the resolution of seismic data. Using the inversion results,it is readily to predict the sand distribution of thin interbedded reservoirs of lower sand section. It is indicates that curve frequency division quasi-acoustic logging inversion can improve vertical and lateral seismic resolution, which makes the prediction of thin interbedded reservoir in complex geology setting more easily.