Research And Application Of The Seismic Attribute Optimization Based On Improved SLLE

Predicting oil and gas by using seismic attributes has been an important methodfor recognization and monitoring on oil-gas reservoir. However, for the complicatedrelationship between seismic attributes and lithology, fluid nature and reservoirpetrography, too many seismic attribute parameters can be extracted from seismic data,result in the redundant and waste seismic data informations. After the optimumprocessiong of seismic attributes can markedly improve precision of reservoirprediction, and also further effectively descript the reservoir. So, it has importantsignificance research the optimum methods of seismic attributes and determine theoptimal seismic attribute parameters.In this paper, the category relationship between attribute sample data is added inthe extracted seismic attribute parameters by fuzzy cluster based on equivalencerelationship at first. Then, dimension and optimization the correlation attributes basedon improved supervised locally linear embedding method. Using after optimizatedcomprehensive seismic attribute parameters to predicate reservoir parameter, such asporosity in order to test the optimal results. At the same time, the paper discussing theparameter choice problem of the algorithm in detail. For judging the method isefficient, factor analysis is used to reduce dimension of the same seismic attributeparameters.The paper first indicated that the seismic attribute optimization base on improvedsupervised locally linear embedding. The algorithm of supervised locally linear is anew non-linear dimension reduction method. It can make the dimension reductiondata consistent with the original topology configuration, and the method has merits ofboth linear and non-linear. Supervised locally linear embedding is used to optimizethe seismic attributes, which fits characteristics of geologic body. The characteristicsinclude the complexity of geological structure, high non-linear of seismic attributesand reservoirs, and high dimension of seismic data. The application is either thecreation of the seismic attribute optimization, or the application domain expansion ofthe supervised locally linear embedding algorithm. Supervised locally linearembedding algorithm requires the original data samples have classifiable informationbesides. Considering the deficiency of transcendent knowledge of the information, thecomplexity and high nonlinear of the seismic attribute data, and they are also fuzzy and incomplete, the fuzzy cluster based on equivalence relationship is put forward forthe classification of the seismic attribute samples.This article applied the algorithm of supervised locally linear into the seismicattribute optimization of horizon and section of some work area. It reduced to fewseveral synthetical attribute parameters from original many seismic attributeparameters, so it achieved reducing dimension and optimization of seismic attributeparameters. And it respectively predicted the values of porosity which correspond tosynthetical attribute parameters which had been gained through SLLE algorithm andFactor Analysis technology and seismic attribute parameters which had not beenoptimized. The results are considered satisfying trough the figs and the analysis oferror. It indicates that applying the improved algorithm of supervised locally linearinto the seismic attribute optimization is feasible.