The Application Of Multiple Information Fusion Method In Seismic Attribute Reservoir Prediction

The carbonate reservoir of TaHe oilfield is deeply buried, generally greater than 5300 m, and bedrocks do not have reservoir properties, reservoir types are mainly cracks and stalactite caves. In the previous studies of TH10, mainly about the Karst weathering crust karst and part of the vertical flow zone segment for the range, from 0 to 80 m, below the T74, and the deeper karsts and seam caves are rarely involved. However, drilling and other datas suggest that the understanding oil seam cave sector still exist the exploitable cave type reservoirs, which is with some evaluation of the potential for development in the current research. Therefore, to increase reservoir forecasting research in the deep cave can deepen understanding of karst development regularity and expand the areas of reserves in the deep reservoir of resources, which can provide geological basis for the transformation of the building positions in TaHe Oilfield.The seismic attribute analysis technology is widely used in reservoir prediction, while the conventional seismic attributes have poor effect and low accuracy, and the results are often multi- solution. However, the information fusion method can solve the contradiction of the correlation between multiple attributes, and the fusion attribute is more accurate and more effective than the single attribute data. Therefore, it is important to detect reservoir cave body and to solve the problem of multi- solution by applying the information fusion technology to seismic attribute reservoir prediction.This thesis mainly studied basic principles and calculation steps of four kinds of information fusion algorithms, including principal component analysis, kernel principal component analysis, fuzzy c-means clustering, kernel fuzzy c-means clustering, and applied these information fusion algorithms to the detection of "beaded" cave reservoirs of TH10 Middle-depth. This thesis mainly obtained three aspects of research results, including methods, software, applications.(1) Methods results. PCA is a good data fusion method, essentially is a linear method, and is applicable to the strong correlation attributes; KPCA’s nonlinear processing ability is stronger than PCA, the information fusion effect also has better detection capability. FCM is a kind of fusion method based on fuzzy mathematical, it’s nonlinear processing ability is not strong; for the attribute datas whose nonlinear relationship is strong, KFCM can the problem that linear space cannot be linear segmentation, so KFCM’s fusion effect is better.(2) Software results. Mainly based on t visual c++ 6.0 platform, this thesis carries on the codes of the four kinds of information fusion algorithms for the seismic attribute datas. Function modules have good effect in reservoir forecasting, and have some application development value.(3) Application results. The application results have three parts, including the geological model, the well profile, and the research area plane:(a) Results of the geological model. This thesis set up a variety of different Stalactite cave geological models to analyze the seismic response characteristics by simulate forward. When the size of the Stalactite cave is too small, seismic waves can’t identify the Stalactite cave. With the increase of the size of Stalactite cave, the response strength of the seismic amplitude is also increased. When the rate of the filling is decreased, the strength of the reflecting seismic response is increasing. When increases at intervals of the Stalactite cave in the case of longitudinal distinguished, the earthquake response intensity is increases.(b) Application results of the well profile. Applying the fusion methods to the well profile, it can be known that the information fusion parameters can detect the location and scale of the cave accurately, and can detect the fluid properties of the karst cave in a certain degree. From the fusion effect, the kernel principal component analysis is stronger than the principal component analysis, and the kernel fuzzy clustering is stronger than the fuzzy clustering.(c) Application results of the research area plane. In order to study TH10 Yingshan Sec, this thesis extracted some attribute parameters by fine horizon tracking, and obtained the caves plane distribution characteristics of the target layer after using information fusion method: The fused plane has the scattered distribution as short axis, overall is "East more west less, north more south less" feature. Integrated production developments suggested that the region which has a medium fusion detection value, and in which caves is centralized and the plane is not communication is the favorable reservoir and possible oil-rich and gas-rich region.In order to detect the reservoir caves of the study area, the thesis usesd multiple information fusion method for the first time, and achieved good prediction. With some innovative, it has some significance for actual production.