Abstract
Frequently repeated 4D seismic has inspired new ways to interpret time-lapse changes. One such application is well2seis, in which 4D signals are quantitatively correlated to well production to create a localized signal that reveals spatial connectivity and fluid pathways in the reservoir. However, the causal relation between multiple seismic vintages and production data in well2seis is strongly dependent on the availability of cumulative fluid volumes that correspond to multiple acquisition periods. Also, it is difficult to disentangle overlapping responses caused by fluctuations from multiple wells. In this study, we propose a machine learning implementation of this technique (well2seisML) to resolve these issues. The new solution is applied to CO2 plume evolution during injection in the Sleipner field. This allows us to optimally integrate the data from injected volumes with the repeat seismic data.
Original language | English |
---|---|
Pages (from-to) | 994-998 |
Number of pages | 5 |
Journal | SEG Technical Program Expanded Abstracts |
Volume | 2023 |
DOIs | |
Publication status | Published - 14 Dec 2023 |
Event | 3rd International Meeting for Applied Geoscience and Energy 2023 - Houston, United States Duration: 28 Aug 2023 → 1 Sept 2023 |
ASJC Scopus subject areas
- Geotechnical Engineering and Engineering Geology
- Geophysics