The Research On Methods Of Reservoir Feature Parameters Estimation And Fuzzy Evaluation

With development of seismic data acquisition and processing techniques,it provides well-conditioned pre-and post-stack seismic volumes.How to effectively extract information of stratigraphic lithology and fluids implicated in seismic data becomes the difficult and hot issue on research of reservoir geophysics.Taken the information in well logs as an oriented data,it has a significant study importance and industrial value to study a method to extract target feature information from seismic data and evaluate them.Therefore,goal of the research is to establish a method and series of techniques of seismic attributes extraction and optimization,reservoir feature parameters' estimation and evaluation,which aims at effectively extracting reservoir feature parameters that could be used to characterize reservoir's lithology,physical property and oil-and gas-bearing property from seismic data,evaluating their availability and credibility,and realizing a quantitative analysis for target reservoirs.Based on systematically studying theory of pre-and post-stack seismic attributes and analyzing their extraction approaches,the study introduces hierarchical clustering scheme(HCS)and partial least square regression(PLSR)method to optimize pre-and post-stack seismic attributes that well matched with information of lithology,physical property and fluid-bearing of reservoir.Based on theoretical study of well logs interpretation and analysis of well logs,fuzzy clustering method is brought to choose out well logs of gamma ray(GR)log,density log(DEN),sonic interval transmit time(? t)log and porosity(POR)log,which are the most important basic parameters for lithological interpretation and could best character reservoirs,as targets of reservoir feature parameters calculation.Based on studying theory and method of relevance vector regression,by model data,it shows advantages of relevance vector regression on regression and prediction.Then,preprocessing on the optimization seismic attributes data and target well logs,and constructing nonlinear relevance vector regression model between the attributes data and target well logs.Finally,the 3D volumes of reservoir feature parameters attain.Based on studying theory and method of fuzzy inference system and comprehensively analyzing reservoir feature parameters and geological data of core,according to principle of lithology interpretation,Mamdani fuzzy inference system is constructed.Meanwhile,it depicts how to choose the parameters and cut fuzzy rule base,and uses the steepest descent algorithm to search the best local weights quotients.Then,in accordance with the optimized fuzzy rule base,it combines reservoir feature parameters and cores to interpreting lithology of reservoirs.As to reservoir classification,it constructs fuzzy inference system based on Takagi-Sugeno(T-S)model,and fuses reservoir feature parameters.Finally,according to its scores,it obtains reservoir classification by linguistic variables output model.Through test on model data and application to field data of Chengbei block of Shengli oilfield,the HCS-PLSR method can effectively choose out the best and less number of the seismic attributes for calculation.Result of relevance vector regression can well character the structural configuration of surfaces and sedimentary system,and well match with the well logs.Mamdani fuzzy inference system can well reveal lithology of reservoirs,and fuzzy inference system based on T-S model can well reflect characteristics of sand bodies' distribution and match with the drilling results.