Abstract
Facies and fracture network modeling need robust, realistic and multi scale methods that can extract and reproduce complex relations in geological structures. Multi Point Statistic (MPS) algorithms can be used to model these high order relations from a visually and statistically explicit model, a training image. FILTERSIM as a pattern based MPS method attracts much attention. It decreases the complexity of computation, accelerates search process and increases CPU per-formance compare to other MPS methods by transferring training image patterns to a lower dimensional space. The results quality is not however as satisfactory. This work presents an improved version of FILTERSIM in which pattern extraction, persisting and pasting steps are modified to enhance visual quality and structures continuity in the realiza-tions. Examples shown in this paper give visual appealing results for the reconstruction of stationary complex struc-tures.
Highlights
Static modeling of complex reservoirs calls for more than two point statistics
Modifications are made to the original FILTERSIM algorithm in several ways of optimizing template size, considering pattern frequency component in database and pattern distance function ranking in candidate bin instead of random selection
Elbows are around the optimum template size which is consistent with the superior results of these sizes in Figures 6 to 7
Summary
Static modeling of complex reservoirs calls for more than two point statistics. Certain features of these objects such as curvilinearity cannot be expressed via two-point relations [1]. Multiple point statistics (observed patterns) of a certain size of training images are stored in a tree structure; Node properties of realizations are assigned one by one in a search process loop. Pattern to pattern methods such as SIMPAT [1] FILTERSIM [16] and DisPAT [17] are completely isolated from two-point statistics and eliminate the probabilistic paradigm in MPS algorithms. These methods inherently take the probabilities of the whole multiple point patterns conditioned to the same multiple point data event from the training image [18]. Computation cost increases marginally compare to the original method, results quality is far better in our Modified FILTERSIM approach
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.