| Autor: |
A. Buniak, J. Zacharski, G.V. Dahl, H.G. Borgos, R. Castañeda-Aguilar, D. Krawiec, R. Suarez-Rivera, W. Prugar, D.A. Handwerger |
| Rok vydání: |
2015 |
| Předmět: |
|
| Zdroj: |
Proceedings of the 3rd Unconventional Resources Technology Conference. |
| DOI: |
10.15530/urtec-2015-2179307 |
| Popis: |
For many unconventional reservoirs, interpretation of subsurface geology alone can be inadequate for determining reservoir potential. As a result, seismic data have not been as fully utilized as they could be. This paper presents a study that leverages multivariate classification methods to integrate core, log and seismic data for the propagation of material properties across two intersecting 2D seismic lines in the Lublin Basin, Poland. In this study, classification of wireline logs is used to isolate zones of expected similar or different material properties. These classes are then integrated with core data in order to determine which classes are most favorable from both a reservoir quality and completion quality perspective (ability to stimulate), based on the large assortment of core measurements. Following characterization of the classification model at log scale, the collection of pre-stack inverted seismic attributes from each 2D line was trained at the near-well traces to recognize the classes provided by an upscaled version of the log-derived classification model. Once the training rules were determined, they were applied to the full 2D data in order to propagate the classes away from the wellbore, and by extension the estimation of all the material properties integrated with the classes from the core measurements. Within the study region, one training well near the intersection of two 2D lines was used. This training well was on the down-thrown block of a large fault identified on one of the lines. Application of the training rules to the seismic attributes on the up thrown block across the fault shows that the up thrown block likely contains a new set of classes (and hence new and uncharacterized material properties) largely different from those available to train the model from the single available well and core. This integrated approach also showed that there is significant vertical and lateral heterogeneity in this portion of the Lublin Basin and that multivariate classification can be an effective means to integrate, scale and propagate material properties through a basin, even with limited available seismic data. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
