Research On Prediction Of Reservoir Fluid Mobility

As the oil industry shifts from increasing production to value optimization,the oil industry pays more and more attention to reservoir quality prediction.Because the percolation ability of the reservoir reflects the quality of the reservoir,the seismic parameters which can reflect the percolation ability of the reservoir are more favored by the seismic interpreter when evaluating the high-quality reservoir.Reservoir fluid mobility,defined as the ratio of rock permeability to fluid viscosity,is a seismic attribute that can be applied to reservoir physical property and permeability evaluation.However,with the change of reservoir structure from simple structure to complex structure,the application of traditional computing methods of reservoir fluid mobility to evaluate high-quality reservoir is restricted by the following four problems:(1)the imaging resolution is insufficient;(2)the attribute is easily affected by strong reflection layer in the process of calculating;(3)the extraction results are not accurate enough to reflect the internal situation of the formation;(4)the prediction results have multiple solutions.To solve the above problems,this paper first introduces the traditional extraction method of reservoir fluid mobility.Then we improve the accuracy of the traditional extraction method by combining with a high resolution time-frequency analysis method to identify thin interbedded reservoirs.Secondly,through the study on two-phase media theory and the inversion algorithm based on sparse constraint,the approximate inversion method of reservoir fluid mobility is proposed.Finally,we use the multi-attribute fusion method based on machine learning to reduce the multi solution of prediction results.The main research work and innovation of this paper are as follows:(1)Reservoir fluid mobility extraction based on time-frequency spectral decomposition.This paper introduces the traditional extraction method of reservoir fluid mobility based on time-frequency analysis method.Aiming at the problem that the existing oil-gas imaging resolution of reservoir fluid mobility is insufficient,a new reservoir fluid mobility extraction method is proposed,and the derivation of the method is described.In this paper,we first develop a high-resolution time-frequency analysis method,the synchrosqueezing generalized S-transform with the Lucy-Richardson algorithm.Secondly,on the basis of this time-frequency analysis method,the large-scale reservoir fluid mobility in the exploration area is calculated.In the process of extracting reservoir fluid mobility,we analyze the difference in the resolution of the attribute sections based on different time-frequency analysis methods,which will affect the accuracy of reservoir prediction.In addition,the Gassmann equation is used to replace the fluid types in the well,and the influence of fluid types on fluid mobility is discussed.(2)Reservoir fluid mobility extraction from the seismic data with strong reflection backgrounds.To solve the problem that the weak reflection reservoir is shielded by the strong reflection backgrounds in extracting of reservoir fluid mobility from seismic data,a work flow of reservoir fluid mobility extraction based on the basis pursuit algorithm is proposed.In this paper,taking buried hill reservoirs as an example,a typical buried hill reservoir model is established.Through the numerical simulation of buried hill reservoir model,the interference of strong reflection background on reservoir prediction is shown.Using basis pursuit to separate the strong reflection layers from seismic data,then the extracted reservoir fluid mobility can reflect the percolation information of the target layer more truly.This method can effectively predict the weak reflection reservoir.(3)Inversion of reservoir fluid mobility from the frequency-dependent seismic data.The two-phase media theory and the inversion theory based on sparse norm constraint are studied.Considering that the location of high fluid mobility obtained by the extraction method is close to the reservoir interface.To obtain the fluid mobility in the middle of the reservoir,an approximate inversion method of reservoir fluid mobility from frequency-dependent seismic data is proposed.Firstly,we calculate the reservoir fluid mobility coefficient using well data according to the relationship of fluid parameters.Then,we establish an inversion equation based on the low-frequency reflection coefficient and the reservoir fluid mobility.Taking the reservoir fluid mobility coefficient calculated from well data as a priori constraint,the low-frequency model is subsequently constructed and applied with the inversion equation to obtain an inversion objective function.Next,the inversion equation is solved by the basis pursuit algorithm.Finally,the proposed reservoir fluid mobility inversion method is applied to synthetic and real data of gas-bearing reservoirs.The real data processing results show that the proposed reservoir fluid mobility inversion method can estimate the fluid mobility in the actual position of the reservoir more effectively.(4)The reduction of the multiple solutions of reservoir fluid mobility prediction results.Reservoir prediction results are often affected by lithology,structure,fluid and other factors.To further reduce the multiple solutions of prediction results,this paper studies the machine learning method and preliminarily explores the application of artificial intelligence in reservoir prediction.By optimizing multiple inversion parameters and sensitive seismic attributes of reservoir fluid,unsupervised machine learning algorithm and supervised machine learning algorithm are used to automatically classify and fuse the optimized data volumes to determine the most favorable reservoirs in the exploration area,which improves the efficiency of seismic data interpretation.