Formation water geochemistry for carbonate reservoirs in Ordos basin, China: Implications for hydrocarbon preservation by machine learning

Hongyan Yu, Zhenliang Wang, Reza Rezaee, Yihuai Zhang, Lezorgia N. Nwidee, Xi Liu, Michael Verrall, Stefan Iglauer

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Formation water can in principal be used to identify hydrocarbon reserves. One such potential reserves are the gas reservoirs in the Ordos basin in China. However, there is limited data for this basin; we thus investigated the geochemical properties of a large range of formation water acquired from the Ordovician in the Ordos basin (42 brine samples obtained from different wells at M5 member) and analyzed their chemical characteristics. The results showed that this formation water is associated with a sealed reservoir, which is good for hydrocarbon storage. This is also related to the demonstrated strong diagenetic transformations. We also proposed statistical relationships between these geochemical properties and hydrocarbon storage based on a machine learning method (Decision tree). The results suggest that the salinity, Na +/Cl ratio, (Cl -Na +)/Mg 2+ ratio, (HCO 3 -CO 3 2-)/Ca 2+ ratio and Mg 2+/Ca 2+ ratio highly correlate with the gas preservation. The results thus provide drastically more accurate predictions in terms of where to find gas reservoirs in the Ordos basin, and can thus lead to significantly better exploitation of these resources.

Original languageEnglish
Article number106673
JournalJournal of Petroleum Science and Engineering
Early online date9 Nov 2019
DOIs
Publication statusE-pub ahead of print - 9 Nov 2019

Keywords

  • Decision tree
  • Formation water
  • Hydrocarbon preservation
  • Ion concentration
  • Machine learning

ASJC Scopus subject areas

  • Fuel Technology
  • Geotechnical Engineering and Engineering Geology

Fingerprint Dive into the research topics of 'Formation water geochemistry for carbonate reservoirs in Ordos basin, China: Implications for hydrocarbon preservation by machine learning'. Together they form a unique fingerprint.

  • Cite this