Skip to the content.

Interdisciplinary Machine Learning in Science and Engineering ()

Eduardo Gildin, Harold Vance Department of Petroleum Engineering
Applications of Machine Learning to Life-Cycle Reservoir Engineering: From Drilling to Reservoir Simulation


In this talk, I present two applications of machine learning to petroleum engineering. I will start with drilling engineering where the objective is to design optimal drilling campaigns by optimizing drilling parameters such as rate of penetration (ROP), weigh on bit (WOB), and drilling bit rotation (RPM) despite dysfunctions (perturbations) encountered due to unknown formation characteristics. I will show how machine learning can help us detecting these dysfunctions in real-time. The second application is in reducing the computational cost of reservoir simulation. I will show how model reduction enhanced by machine learning can mitigate the computation cost associate with the large scale optimization framework in optimal reservoir production and well placement. Although both applications seem disjoint, they are in fact, connected by the so-called closed-loop reservoir management, whereby selecting well locations and performing optimal drilling campaigns can enhance the life-cycle, and in turn, optimal hydrocarbons production of a particular reservoir. They can be seen as subsystems in closed-loop feedback loops associated with control of reservoirs and drilling automation.