TUPREP is a cooperative industry/university research project which was organized at the end of 1988. Our mission is to conduct fundamental and applied research and development in the areas of well testing, reservoir characterization and reservoir simulation. TUPREP formally began operation on January 1, 1989. Prior to our First advisory board meeting in May of 1989, we had received formal commitments for 1989 membership from four companies. Subsequent to 1989, the number of member companies has fluctuated reaching a high of eleven in 1995 and a low of four in 2000.
Although our ultimate goal is to provide techniques for practical application, some of our research is fundamental in nature, as our objective is to develop methodology that has a sound theoretical basis. Our primary objective is to provide new insights, useful information and tools to our member companies, but we are also dedicated to equipping graduate students with up-to-date technical knowledge and skills so they can become productive engineers in a research and development or technical service environment. To date, 49 Ph.D. dissertations and 16 M.S. projects have been completed with support from TUPREP and delivered to member companies. In addition, over sixty papers have been written for publication based on TUPREP research.
The mission of TUPREP is to develop and implement novel technology for reservoir optimization, assisted history matching and uncertainty quantification that is theoretically sound and feasible for practical application for both conventional and unconventional reservoirs.
Recent/current TUPREP research projects, some of which contain multiple subprojects, include the following:
- Nonlinearly constrained single- and multi-objective production optimization.
- Deep-learning proxy-based reservoir prediction and production optimization.
- Physics-based data-driven interwell waterflooding simulator.
- Ensemble-based methods including ES-MDA.
- Three-dimensional physics-based data-driven models for history matching, characterization and monitoring of waterflooding and for waterflooding optimization.
- Machine learning workflows for history-matching, production optimization and uncertainty quantification.
- Reservoir optimization using a new support vector regression (SVR) implementation with adaptive training.
- Analytical estimation of optimal hyper-parameters for SVR.
- Generating proposal distributions to optimize computational efficiency of Markov chain Monte Carlo.
- Using reservoir physics to improve the accuracy of stochastic gradients for production optimization with nonlinear constraints.
- Optimization of well trajectories and controls.
- Multiobjective reservoir optimization.
- Adjoint method for gradient-based history matching and life-cycle production optimization.
- Derivative-free methods for optimal well controls.
- Semi-analytical solutions for two-phase flow injection/falloff/production testing.
- Integrating pressure transient data into reservoir models.