Stochastic Inversion Of Reservoir Parameters Based On Gradual Deformation Method

Heterogeneity in real earth medium plays an important role in the study of oil and gas reservoir,metallic ore deposit,and hydrology.Heterogeneity is associated with different sedimentary environments and tectonic movements.And it indicates grain size,sequence and continuity.It is significant to recover heterogeneity from seismic data,well log data and geology.Seismic inversion is an important method to obtain heterogeneity in real earth medium.It is impossible to get an accurate reservoir model due to the limited bandwidth of seismic data.Additional constraints from geology(e.g.sedimentary facies and structure),well data and spatial correlation of medium could provide important information.Geostatistical inversion provides a better strategy for generating a suite of alternative subsurface property realizations conditioned on different kinds of data and spatial correlations.Modeling and optimization algorithms are the keys to successful application of geostatistical inversion.In this paper,a new stochastic seismic inversion method based on the local gradual deformation method is proposed to invert elastic and rock-physics properties.This approach can incorporate seismic data,well data,geology and their spatial correlations into the inversion process.Indicator cokriging,probability field simulation and FFT moving average generator could build reservoir models conformed to geology,well data and spatial correlations.Gradual deformation method could guarantee all the models we build and modify constrained by a priori information in the inversion process.We can apply the methodology many times using different random seeds and then evaluate the uncertainty.Three improved strategies are proposed to make original gradual deformation method suitable for seismic inversions.The improved modeling and inversion method can be used in acoustic inversion based on post-stack seismic data,elastic inversion based on pre-stack seismic data and simultaneous inversion.The applications to several synthetic examples and real cases study demonstrate that our approach is effective and feasible.The main contents are as follows:(1)The study of stochastic modeling method,which could integrate different kinds of data such as geology and well data;(2)Seismic inversion based on gradual deformation method and its improved strategies;(3)The GPU parallelization of stochastic modeling and gradual deformation method;(4)Inversion of elastic and rock-physics properties using stochastic modeling,gradual deformation method and rock-physics model.By research of this paper,the results and conclusions are as follows:(1)FFT moving average generator draws the randomness in the spatial domain,which allows us to uncouple the random numbers from the structural parameters.this feature constructs the link between random numbers and model in spatial domain so that it can be used in sub-regional modeling and local modification.Modeling by indicator cokriging,FFT moving average generator and probability field simulation could generate models satisfied with geology,well data,any probability distribution and spatial correlations;(2)Gradual deformation method updates the Gaussian white noise rather than the model,which could guarantee all the models we build and modify in inversion process conformed to geology,well data and spatial correlations.The efficiency of our proposed method improves significantly.(3)The parallelization of FFT moving average generator and gradual deformation method is easier than the inversion using sequential simulation.(4)The applications in synthetic and real cases show that our method could integrate seismic data,well log data,geological information and their spatial variability.In addition,our method combines the advantage of deterministic inversion and stochastic inversion,which means it has high efficiency,generates reasonable results and could give uncertainty estimations.This dissertation is mainly divided into five chapters.The first chapter is a brief introduction to the research background,purpose,significance,research progress status at home and abroad,as well as the thesis' s main research content,technical route and innovations.The second chapter introduces the basic theory of FFT moving average method,indicator cokriging,probability field simulation and the advantages of their combination.Some experiments are shown to demonstrate such advantages.The third chapter discusses the basic theory of gradual deformation method,three improved strategies we proposed based on the characteristic of seismic data,and the GPU parallelization of stochastic modeling and gradual deformation method.This chapter also discusses the applications of gradual deformation method in acoustic inversion based on post-stack seismic data,elastic inversion based on pre-stack seismic data and simultaneous inversion.In chapter four,three real cases are shown to test our improved gradual deformation method.The last chapter summarizes the thesis' s main work and conclusions and puts forward the future research directions.