Name
Title | ||
|---|---|---|
A training image evaluation and selection method based on minimum data event distance for multiple-point geostatistics. |
Abstract | ||
|---|---|---|
A training image (TI) can be regarded as a database of spatial structures and their low to higher order statistics used in multiple-point geostatistics (MPS) simulation. Presently, there are a number of methods to construct a series of candidate TIs (CTIs) for MPS simulation based on a modeler's subjective criteria. The spatial structures of TIs are often various, meaning that the compatibilities of different CTIs with the conditioning data are different. Therefore, evaluation and optimal selection of CTIs before MPS simulation is essential. This paper proposes a CTI evaluation and optimal selection method based on minimum data event distance (MDevD). In the proposed method, a set of MDevD properties are established through calculation of the MDevD of conditioning data events in each CTI. Then, CTIs are evaluated and ranked according to the mean value and variance of the MDevD properties. The smaller the mean value and variance of an MDevD property are, the more compatible the corresponding CTI is with the conditioning data. In addition, data events with low compatibility in the conditioning data grid can be located to help modelers select a set of complementary CTIs for MPS simulation. The MDevD property can also help to narrow the range of the distance threshold for MPS simulation. The proposed method was evaluated using three examples: a 2D categorical example, a 2D continuous example, and an actual 3D oil reservoir case study. To illustrate the method, a C++ implementation of the method is attached to the paper. Candidate training image (CTI) evaluation and optimal selection method is proposed.The proposed method is based on minimum data event distance (MDevD).It evaluates conditional data and CTI compatibility for multiple-point statistics.Three examples illustrating its efficacy are also presented. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1016/j.cageo.2017.04.004 | Computers & Geosciences |
Keywords | Field | DocType |
Multiple-point geostatistics,Candidate training images,Evaluation and optimal selection,Minimum data event distance | Data mining,Compatibility (mechanics),Ranking,Mean value,Computer science,Categorical variable,Data grid,Higher-order statistics,Statistics,Geostatistics | Journal |
Volume | Issue | ISSN |
104 | C | 0098-3004 |
Citations | PageRank | References |
0 | 0.34 | 5 |
Authors | ||
5 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Wenjie Feng | 1 | 2 | 2.03 |
| Shenghe Wu | 2 | 0 | 0.34 |
| Yanshu Yin | 3 | 0 | 0.34 |
| Jiajia Zhang | 4 | 26 | 16.64 |
| Ke Zhang | 5 | 75 | 21.74 |