Loading...

A sensitivity study of FILTERSIM algorithm when applied to DFN modeling

Ahmadi, R ; Sharif University of Technology

751 Viewed
  1. Type of Document: Article
  2. DOI: 10.1007/s13202-014-0107-0
  3. Abstract:
  4. Realistic description of fractured reservoirs demands primarily for a comprehensive understanding of fracture networks and their geometry including various individual fracture parameters as well as network connectivities. Newly developed multiple-point geostatistical simulation methods like SIMPAT and FILTERSIM are able to model connectivity and complexity of fracture networks more effectively than traditional variogrambased methods. This approach is therefore adopted to be used in this paper. Among the multiple-point statistics algorithms, FILTERSIM has the priority of less computational effort than does SIMPAT by applying filters and modern dimensionality reduction techniques to the patterns extracted from the training image. Clustering is also performed to group identical patterns in separate partitions prior to simulation phase. Various practices including principal component analysis, discrete cosine transform and different data summarizers are used in this paper to investigate a suitable way of reducing pattern dimensions using outcrop maps as training image. Because of non-linear nature of patterns present in fracture networks, linear clustering algorithms fail in determining the borders between the actual partitions; non-linear approaches like Spectral methods, however, can act more efficiently in diagnosing the right clusters. A complete sensitivity analysis is performed on FILTERSIM algorithm regarding search template dimension, type of filtering technique and the number of clusters for each clustering approach described above. Interesting results are obtained for each parameter that is changed during analysis
  5. Keywords:
  6. Algorithms ; Complex networks ; Discrete cosine transforms ; Fracture ; Principal component analysis ; Reduction ; Sensitivity analysis ; Clustering ; Dimensionality reduction techniques ; Discrete fracture network ; Geometrical modeling ; Multiple-point geostatistics ; Clustering algorithms
  7. Source: Journal of Petroleum Exploration and Production Technology ; Vol. 4, issue. 2 , June , 2014 , p. 153-174 ; ISSN: 21900558
  8. URL: http://link.springer.com/article/10.1007%2Fs13202-014-0107-0