Research On Reservoir Modeling Method Based On Multi-point Geostatistics

Reservoir is the core content in the field of oil and gas exploration.Accurate identification and positioning of the reservoir is directly related to the overall revenue of the oil and gas field.Among them,the prediction of reservoir lithofacies and res ervoir physical parameters is the most important,and it is an important content of reservoir description and reservoir evaluation.Reservoir characterization is a key method to obtain the reservoir facies and its physical properties.How to accurately characterize the reservoir becomes a challenge.Multi-point geostatistics is a new reservoir simulation method emerging in recent years.One of its advantages is that it can combine multiple information to describe the reservoir,that is,joint well data and seismic data.Conventional reservoir characterization based on well data is difficult to constrain the area between wells.Therefore,it is necessary to introduce seismic information in the simulation process.The key issue is how to obtain credible,high-precision seismic information and how to integrate seismic information into the simulation process.This paper first studies how to obtain high-precision and reliable seismic information.Aiming at the shortcomings of the traditional Bayesian method,a probabilistic neural network(PNN)is proposed.This method can not only obtain reliable prediction results,but also obtain the posterior probability of the prediction results.This posterior probability can be used not only as the basis for the uncertainty evaluation of the results,but also can be integrated into the multi-point geological statistics as the original data During the modeling process,constraints are provided for the simulation of the interwell area.The core of the thesis is reservoir prediction.This article compares the tw o-point and multi-point geostatistics,and selects the multi-point geostatistics method as a tool.Multipoint geostatistics uses training images as a tool to characterize underground reservoirs by depicting the relationships between multiple points in space.Compared to the classic SNESIM algorithm,the IMPALA algorithm uses a list instead of a search tree,uses less storage space,and allows for multi-phase simulation.Multi-grid technology is used to realize multi-scale structure simulation,and Tau model method is used to realize well-seismic joint constraint.Compared with single well constraint,the stability of simulation results is greatly enhanced and randomness is reduced...