Loading...
Geometrical Fracture Modeling Within Multiple-Point Statistics Framework
Ahmadi, Rouhollah | 2011
924
Viewed
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 41596 (06)
- University: Sharif University of Technology
- Department: Chemical and Petroleum Engineering
- Advisor(s): Masihi, Mohsen; Rasaei, Mohammad Reza; Eskandaridalvand, Kiomars; Shahalipour, Reza
- Abstract:
- Majority of the oil and gas reservoirs, in the main hydrocarbon production regions around the world, are naturally fractured reservoirs. Fractures play an important role in reservoir fluid flow either in the form of high permeable complex conduits or strong permeability anisotropies. Realistic characterization of naturally fractured reservoirs requires an exhaustive understanding of fracture connectivity and fracture pattern geometry. These subsequently demand description of many fracture parameters such as density (intensity), spacing, orientation, size and aperture. Therefore, a first step in fractured reservoirs characterization is the static geometric modeling of the subsurface fracture system. The large number of unknowns and great amount of uncertainties about subsurface fractures, however, may not allow a single deterministic representation. In contrast to the deterministic approaches, this thesis presents stochastic geometrical modeling of naturally fractured reservoirs within Multiple Point Statistics (MPS) framework.There are two major strategies to model fracture patterns stochastically; in object-based algorithms fractures are represented as randomly but independently distributed planar convex objects defined by their centroid, shape, size and orientation which in most cases do not provide realistic realizations. It is not possible consequently to depict the undulation and curvature of fractures mainly due to the structural and mechanical heterogeneities of the reservoir. Constructed realizations from pixel based algorithms, on the other hand, are built by simulating one pixel at a time, thus offering great flexibility for constraining to data of diverse supports and types. However application of traditional two-point statistics, even accompanying the pixel based approaches, may not be adequate in capturing the complex curvilinear geological features like tectonic fractures. Newly developed multiple-point geostatistical algorithms, including pattern-pasting based simulations like SIMPAT and FILTERSIM can be effectively utilized in representing complex fracture patterns.This study starts by a comprehensive investigation on the available MPS algorithms and goes on to provide some remedies in removing the simulation challenges; It’s possible to efficiently reproduce the structural patterns present in the training images using a newly developed dynamic approach to determine the optimal size of the scanning template. In addition, defining a new intermediate space using the artificial intelligence tools like as the fuzzy inference systems and neural networks helps to improve the data conditioning of the MPS algorithms. Finally, various dimensionality reduction techniques will be reviewed as in the FILTERSIM algorithm; spatial filters as well as the principal component analysis (PCA) technique could be assimilated effectively in reducing the large number of extracted features before storing in the pattern database. These would be practically represented at the end by implementing the MPS algorithms on the real outcrop images.
- Keywords:
- Dimensionality Reduction ; Fracture Network Modeling ; Multiple Point Statistics ; Training Image ; Patterns Database ; Soft Data
- محتواي پايان نامه
- view