Contribution of Artificial Intelligence and Machine Learning in U.S. DOE's Efforts During the Aftermath of Deepwater Horizon

Abstract

Abstract Petroleum Engineering and Analytics Research Laboratory at West Virginia University (PEARL) played a key role in estimating the amount of oil being discharged into the Gulf of Mexico during the Deepwater Horizon disaster in 2010. PEARL's calculation of the amount of oil that was being discharges into the Gulf of Mexico were based on the Smart Proxy technology. Smart Proxy is a unique an innovative implementation of Artificial Intelligence and Machine Learning to develop fast and accurate proxy models of high fidelity numerical simulations that do not compromise the complex physics and the high resolution of the original numerical simulation models. Smart Proxy modeling was implemented by many of PEARL's scholars for their Ph.D. dissertations. Execution of the Smart Proxy models takes from fractions of a second to few minutes on a PC workstation (laptop) to accurately replicate detail results of high fidelity numerical simulation models that take hours or days to run on High Performance Computational facilities (HPC - clusters of large number of CPUs or GPUs). As a subcontractor to the U.S. Department of Energy, National Energy Technology Laboratory (NETL), PEARL was asked to join a team of scientists and engineers from across the United States from multiple DOE national Labs in order to start an effort to estimate the amount of oil that was being discharged in the Gulf of Mexico. Since the initial BP's estimate of 1,000 barrels per day seemed an under-estimation (this estimate was later increased to 5,000 barrels per day after the initial estimate were challenged by the U.S. government), the Obama administration asked the secretary of Energy (the Noble Laureate, Steven Chu) to set up a team for providing a realistic estimate of the oil discharge rate into the GOM. This article presents a summary the efforts by PEARL in estimating the oil discharge rate into the GOM during the Deepwater Horizon disaster. PEARL's use of Smart Proxy modeling of numerical reservoir simulation models coupled with detail numerical flow in pipe and network models allowed the team to generate millions of scenarios in order to exhaustively examine and quantify the uncertainties associated with all that was unknown and were being estimated at the time of analyses. The final estimate of the oil discharge rate into the GOM calculated by PEARL proved to be highly accurate once the well was contained and the flow was actually measure providing a valuable use-case of the application of the application of Artificial Intelligence and Machine Learning in the upstream oil and gas industry.