Delineation of gas hydrate reservoirs in the Ulleung Basin using unsupervised multi-attribute clustering without well log data
- Authors
- Lee, Jaewook; Byun, Joong moo; Kim, Bona; Yoo, Dong-Geun
- Issue Date
- Oct-2017
- Publisher
- ELSEVIER SCI LTD
- Keywords
- Quantitative seismic interpretation; Multi-attribute analysis; Ulleung Basin; Gas hydrate; Unsupervised clustering; Machine learning
- Citation
- JOURNAL OF NATURAL GAS SCIENCE AND ENGINEERING, v.46, pp.326 - 337
- Indexed
- SCIE
SCOPUS
- Journal Title
- JOURNAL OF NATURAL GAS SCIENCE AND ENGINEERING
- Volume
- 46
- Start Page
- 326
- End Page
- 337
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/151493
- DOI
- 10.1016/j.jngse.2017.08.007
- ISSN
- 1875-5100
- Abstract
- To target a well location for gas hydrate reservoirs, many previous studies have used only some discontinuous indicators such as the bottom simulating reflector (BSR) and dimming or acoustic impedance (AI). However, when drilling the first well, there are no well log and core analysis data for Al inversion and geological analysis. Moreover, these methods may have some risk of finding gas hydrate zones and overestimating the reservoir distribution. First of all, we inverted Al using seismic data and root-mean-square (RMS) velocity for AI inversion excepting well logs. In this inversion, the low frequency impedance variations were estimated from RMS velocity by the product of the interval velocity calculated by the Dix equation and the bulk density from this interval velocity. Then, this low frequency information was integrated with the seismic frequency information obtained from a reflectivity series by sparse-spike impedance inversion. To prevent overestimation due to the use of AI alone, we focused on another rock property, shear impedance (SI), which indicates the 'rigidity' of rocks, as gas hydrate consolidate sediments and increases the value of SI noticeably compared to that in surrounding non-reservoirs. To estimate this property, we used this inverted AI and partial stack seismic data using the two-term Fatti's equation. This amplitude variation with an offset (AVO) equation excludes the density term, which is too sensitive to noise, and has only the two terms of AI and SI. As a result, we applied K-mean clustering, which is the method of unsupervised machine learning, to delineate a more accurate and quantitative distribution of hydrated reservoirs from areas with higher impedances and higher values of two additional attributes (RMS amplitude and instantaneous frequency) compared to surrounding formations. In conclusion, we verified that this workflow is useful for identifying the distribution of potential reservoirs and aiding in well-site location determination.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 공과대학 > 서울 자원환경공학과 > 1. Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.