Data-driven machine learning approach to predict mineralogy of organic-rich shales: An example from Qusaiba Shale, Rub’ al Khali Basin, Saudi Arabia

Ayyaz Mustafa, Zeeshan Tariq, Mohamed Mahmoud, Ahmed E. Radwan, Abdulazeez Abdulraheem, Mohamed Omar Abouelresh

Research output: Contribution to journalArticlepeer-review

33 Scopus citations

Abstract

The Qusaiba Shale is a proven source rock for the Palaeozoic petroleum system of the middle east and it is targeted as a potential source of unconventional shale gas potential reservoirs in Saudi Arabia and recently considered a possible source of unconventional gas. Mineralogy of shale plays a key role in the successful design and performing the hydraulic fracturing operations and in turn evaluating the production potential. There is limited research available in the literature on the application of artificial intelligence for mineralogical prediction, which motivate us to perform this research on the Qusaiba Shale. The study aims to predict the Qusaiba Shale mineralogy, specifically clay and quartz minerals using readily available conventional logs, where both minerals are the major constituents of shale and help in assessing the brittle and ductile zones within the shale formation. Three lithofacies were defined in the Qusaiba Shale based on different geological and sedimentary features of core samples, and they have different ranges of quartz and clay content. Predictive models were developed for two utmost important minerals present in Qusaiba Shale formation including clay and quartz using artificial intelligence techniques. An adaptive neuro-fuzzy inference system (ANFIS) and artificial neural networks (ANN) were employed to precisely predict the two major minerals contents in Qusaiba Shale formation. All four models were found to have good accuracy. ANN-based models exhibited minimal errors with AAPE and RMSE of 2.53 and 1.45 for quartz and 3.43 and 2.01 for unseen data points for clay prediction models respectively. ANFIS-based prediction models presenting the AAPE and RMSE 2.67 and 1.83 for predicted quartz values and 3.59 and 1.89 for the clay content respectively. The applied artificial intelligence predictive model for major minerals in Qusaiba Shale i.e., clay and quartz would be a viable and useful approach to achieve valuable information about minerals content in shale for prospective evaluation of shale gas reservoirs. The results of this work would be useful in prospect evaluation and optimizing the production from unconventional shale gas reservoirs in Saudi Arabia. Moreover, the approach used may provide insights into replacing traditional mineralogy determination methods to save cost and resources in the absence of cores and mineralogy logs in the unconventional Qusaiba Shale resource. Furthermore, applying the same artificial intelligence approach and steps could be used for the field development of other shale gas and oil fields worldwide.
Original languageEnglish (US)
JournalMarine and Petroleum Geology
Volume137
DOIs
StatePublished - Mar 1 2022
Externally publishedYes

Bibliographical note

Generated from Scopus record by KAUST IRTS on 2023-09-20

ASJC Scopus subject areas

  • Economic Geology
  • Oceanography
  • Stratigraphy
  • Geophysics
  • Geology

Fingerprint

Dive into the research topics of 'Data-driven machine learning approach to predict mineralogy of organic-rich shales: An example from Qusaiba Shale, Rub’ al Khali Basin, Saudi Arabia'. Together they form a unique fingerprint.

Cite this