Distinguished Lecture - Subsurface Analytics: Digital Transformation of Reservoir Management with Artificial Intelligence and Machine Learning


April 19th, 2021 by Jasmine Humaira

On April 19th, 2021, SPE Java Section held a distinguished lecture session with Mr. Shahab D. Mohagegh as the speaker with the theme “Subsurface Analytics: Digital Transformation of Reservoir Management with Artificial Intelligence and Machine Learning”. The session lasted for about an hour in which it involves a lecture and Q&A session.

Mr. Mohaghegh is considered a pioneer in the application of Artificial Intelligence (AI) and Machine Learning (ML) in the Exploration and Production industry. Having worked in the field for more than 30 years, his profound understanding of the subject could be seen in his numerous research works—ranging from three of his self-authored books (Shale Analytics, Data-Driven Reservoir Modelling, Application of Data-Driven Analytics for the Geological Storage of CO2) to more than 200 technical papers. He has also been engaged in 60 projects for the independents, NOCs and IOCs and has received an honor from the U.S. Secretary of Energy for his AI-based technical contribution in the aftermath of the Deepwater Horizon incident (Macondo Field, Gulf of Mexico, 2011). Currently, he works as a Professor of Petroleum and Natural Gas Engineering at West Virginia University and as the president and CEO of Intelligent Solutions, Inc. (ISI).

Taking upon the major topic of the lecture with Petroleum Data Analytics (PDA), Mr. Mohagegh priorly highlighted the definition of PDA that is, “PDA is the application of AI and ML in oil and gas industry by using hard data or a real field measurement data—not ones which are generated by the equation—as the main building blocks for its analysis, workflows, modeling and decision making”. This data will be used to model physics by employing Subsurface Analytics (SA), thus, serve conclusions that are needed. By this differing method to conventional numerical reservoir modeling, Mr. Mohagegh explained SA's main advantage which lies on its approach that avoids any assumptions, simplifications, and biases and requiring no mathematical equations. Since both hard data and SA domain experts’ (i.e. geoscientists, reservoir engineers, etc) knowledge are integrated in the method, Mr. Mohagegh furtherly clarified that SA is completely an explainable AI (XAI)—not a black-box as it has often been misunderstood.

Furthermore, the lecture dug deep about SA model applied in conventional resources, the Top-Down Modelling (TDM), from how it works and what it offers. Mr. Mohagegh himself classified the workflow into three simple terms:

  • Learn/Descriptive Analytics: TDM performs intelligent data patching and Well Performance Analysis (WPA).
  • Model/Predictive Analytics: TDM uses AI & ML Algorithms to do a “fully automated history matching”. This unique history matching method employs “Blind Validation in Space” and “Blind Validation in Time” in its process which is advantageous for handling missing data as commonly found in data from mature fields.
  • Optimize/Prescriptive Analytics: TDM gives out uncertainty quantification, well performance prediction, and perform capability in optimizing infill location, choke setting, recovery, and injection location/amount.

In addition, the attendees were also invited to look upon several real case studies as well as feedback from some parties stating forecast with a high percentage of accuracy projected by the TDM model, years after completion of a TDM project.

Looking at the extensive advantages that AI and ML have offered in reservoir management, Mr. Mohagegh closed the lecture with a reminding note that is, “It’s very important to be exposed to this technology. But, remember that AI and ML still need teaching, thus, it’s still very important to be a good domain expertise”.