Title page for ETD etd-06192012-153816


Type of Document Dissertation
Author Shuai, Yuanyuan
URN etd-06192012-153816
Title Strategies for Real Time Reservoir Management
Degree Doctor of Philosophy (Ph.D.)
Department Petroleum Engineering
Advisory Committee
Advisor Name Title
White, Christopher D. Committee Chair
Thompson, Karsten Committee Member
Tyagi, Mayank Committee Member
Zhang, Hongchao Committee Member
Bentley, Sam Dean's Representative
Keywords
  • gathered EnKF
  • ensemble Kalman filter
  • multiscale regularization
  • ensemble based optimization
  • oil price uncertainty
  • bootstrap forecasting
  • sequential Gaussian simulation forecasting
  • production optimization
  • history matching
Date of Defense 2012-05-18
Availability unrestricted
Abstract
Real–time reservoir management is developed to manage a shrinking labor force and rising demand on energy supply. This dissertation seeks good strategies for real–time reservoir management. First, two simulator–independent optimization algorithms are investigated: ensemble–based optimization (EnOpt) and bound optimization by quadratic approximation (BOBYQA). Multiscale regularization is applied to both to find appropriate frequencies for well control adjustment. Second, two gathered EnKF methods are proposed to save computational cost and reduce sampling error: gathered EnKF with a fixed gather size and adaptively gathered EnKF. Finally, oil price uncertainty is forecasted and quantified with three price forecasting models: conventional forecasting, bootstrap forecasting and sequential Gaussian simulation forecasting. The relative effect of oil price and its volatility on the optimization strategies are investigated.

A number of key findings of this dissertation are: (a) if multiscale regularization is not used, EnOpt converges to a higher net present value (NPV) than BOBYQA—even though BOBYQA uses second order Hessian information whereas EnOpt uses first order gradients. BOBYQA performs comparably only if multiscale regularization is used. Multiscale regularization results in a higher optimized NPV with simpler well control strategies and converges in fewer iterations; (b) gathering observations not only reduces the sampling errors but also saves significant amount of computational cost. In addition, adaptively gathered EnKF is superior to gathered EnKF with a fixed gather size when the prior ensemble mean is not near the truth; (c) it is shown that a good oil price forecasting model can improve NPV by more than four percent, and (d) instability in oil prices also causes fluctuation in optimized well controls.

Files
  Filename       Size       Approximate Download Time (Hours:Minutes:Seconds) 
 
 28.8 Modem   56K Modem   ISDN (64 Kb)   ISDN (128 Kb)   Higher-speed Access 
  YuanyuanShuaiDissertation.pdf 13.62 Mb 01:03:02 00:32:25 00:28:21 00:14:10 00:01:12

Browse All Available ETDs by ( Author | Department )

If you have questions or technical problems, please Contact LSU-ETD Support.