The Research Of Reservoir Characteristics Parameter Prediction Based On Seismic Attributes

Pre-stack seismic data contains a wealth of information,such as physical,lithological,fluid properties,etc.It needs a variety of methods to extract the above information which reflects reservoir parameters,and seismic attribute is an important way to obtain these parameters.It can get quite accurate information through the analysis of seismic attributes.Subsurface formation,such as layer thickness,porosity,permeability,saturation characteristics parameters,can only be determined by drilling information,but these various characteristics parameters obtained from the well are only the information of well point location and the above parameters can only determine some discrete points' reservoir information.Therefore,to understand the entire three-dimensional spatial distribution laws of the reservoir characteristics parameters,it is necessary to establish the relationship between seismic data and various reservoir characteristics obtained from well logging data.First,correlation analysis found that,compared to conventional seismic attributes based on post-stack seismic data,seismic attributes extracted on pre-stack angle domain common image gathers(ADCIGs)data contain richer original seismic information.That is,the post-stack seismic data lose some valid faint information of pre-stack seismic traces.Therefore,extracting conventional seismic attributes on ADCIGs,and pre-stack AVO attributes including intercept(P)and gradient(G),then constructing the sample sets together with the resampled well logging data.And standardizing the sample sets with maximum difference normalization method.Furthermore,we proposed GA-SVR seismic attributes selection method which combines the genetic algorithm(GA)and support vector regression(SVR)algorithm,and introduced the tabu search algorithm(TS)into seismic attributes selection.Through the comparison of the GA-SVR and TS method,we chose the GA-SVR method to select the seismic attributes subsets of sample sets.Finally,constructing the non-linear SVR prediction model between the sample sets after seismic attributes selection and well logging data which can reflect the reservoir parameters such as Gamma Ray,Porosity.And then,predicting the characteristics of reservoir parameters of three-dimensional work area.In order to verify the correctness of the proposed method,we tested the method with the F3 demo data provided by Opendtect software and field data of Shengli Oilfield,the estimated results have a high consistency with the actual well logging data at the location of the well,and at the location away from wells,the estimated results could match the seismic data and geological background.