Abstract

In this talk, I present some novel applications of the life-cycle production optimization problem considered under the closed-loop reservoir management studies for conventional and unconventional oil/gas reservoirs. Life-cycle production optimization is a subsequent step of assisted history matching in reservoir management. It can obtain the maximum economic benefit, usually measured as net present value (NPV), by optimizing the well design parameters, which may include wells’ operating conditions such as injection rates and production bottomhole pressures at wells, and well locations at each control step during a reservoir’s lifetime. In this talk, firstly, I present the efficient application of the least-squares support vector (LS-SVR) regression method for optimization of a complex CO₂ Huff-n-Puff Process in unconventional reservoirs in comparison with the conventional optimization method based on the stochastic gradient method (StoSAG). Secondly, I present an efficient gradient-based framework to solve nonlinearly constrained production optimization problems, namely Line-search Sequential Quadratic Programming (SQP) with StoSAG gradients using both high-fidelity and deep-learning proxy methods such as embed-to-control observe. The nonlinearly constrained optimization problems typically refer to optimizing (minimizing) an objective function, subject to some nonlinear state constraints. This type of optimization problem is the focus of most industrial businesses because, in practice, along with bound constraints, there usually exist operational limitations due to the processing capability of the surface facilities, which need to be taken into consideration. Some examples include field liquid production rate (FLPR), field gas production rate (FGPR), water cut (WC), etc. I demonstrate the applicability of this efficient gradient-based production framework on a large-scale three-phase Brugge reservoir model developed by the Dutch Organization for Applied Scientific Research (TNO) as a benchmark case for closed-loop reservoir management.

Biography

Mustafa Onur is the McMan Endowed Chair Professor of the McDougall School of Petroleum Engineering at the University of Tulsa, Director of TU Petroleum Reservoir Exploitation Projects (TUPREP), and Emeritus Professor of Petroleum and Natural Gas Engineering at Istanbul Technical University. Before joining TU, he was a professor at Istanbul Technical University (ITU), Turkey, for 27 years. He also held the Schlumberger Professorial Chair at the Department of Petroleum Engineering at Universiti Teknologi Petronas (UTP), Malaysia. He served as the Department Head of Petroleum and Natural Gas Engineering at ITU from 2006 to 2012 and as the Department Head at McDougal School of Petroleum Engineering at TU from 2016 to 2020. His current research is on the application of inverse problem theory, mathematical optimization, and data science to problems of relevance in optimal reservoir management and development, assisted history matching and uncertainty quantification for oil, gas, pressure, and wireline formation testing, and conventional and unconventional geothermal reservoirs. Onur holds a BS degree from the Middle East Technical University in Turkey, an MS degree, and a Ph.D. degree from the University of Tulsa, all in petroleum engineering. He has served on the editorial committees of SPE Reservoir Evaluation & Engineering and SPE Journal. He is currently the Technical Editor of the SPE Journal and the Associate Editor of the Journal of Petroleum Science and Engineering (continued as Geoenergy Science and Engineering). He is the recipient of the 2010 SPE Formation Evaluation and 2018 Reservoir Description and Dynamics Awards, and a distinguished SPE member since 2014.