Advanced Closed-Loop Reservoir Management for Computationally Efficient Data Assimilation and Real-Time Production Optimization of SAGD Reservoirs

• Author / Creator

  Patel, Rajan G.

• Steam-assisted gravity drainage (SAGD), an in-situ thermal oil recovery method is successfully utilized to extract bitumen from the Canadian oil sands. To improve the reservoir performance, an idea of closed-loop reservoir management (CLRM) was proposed that comprises near-continuous data assimilation to estimate unknown parameters and model-based optimization at distinct decision-making levels. Commercialization of CLRM is vital, however, it poses a significant challenge given the intensive computational requirements of data assimilation techniques, highly nonlinear nature of SAGD process and rare field-scale testing of the concept. In this research, limitations of the fundamental elements of CLRM i.e., data assimilation and short-term production optimization are addressed by developing computationally efficient dynamic modeling workflows and advanced control frameworks. Data assimilation using contemporary techniques at reservoir scale requires high fidelity simulation of many stochastic realizations, resulting in an impractical computational cost. Two solutions proposed in this work are initial sampling method and metamodel. To select few realizations from the ensemble in proposed “scenario reduction” method, Kantorovich distance that quantifies disparity between ensembles using static measures is minimized by solving a mixed-integer linear programming (MILP) problem. In another approach, computationally expensive reservoir simulator is replaced by an integrated Karhunen-Loeve (KL) parameterization and polynomial chaos expansion (PCE) metamodel in Ensemble Kalman Filter (EnKF) and Markov chain Monte Carlo (MCMC). Short-term (or real-time) production optimization using model predictive control (MPC) requires adequate representation of a complex, spatially distributed, nonlinear SAGD process. Variants of MPC that integrate real-time production and temperature data along with well and surface constraints are implemented in this research to achieve steam conformance and an optimum subcool. Adaptive MPC involves continuous re-estimation of model coefficients at each control interval, reflecting the current reservoir dynamics while gain-scheduled MPC decomposes the control problem in a parallel manner with a separate controller for each operating region. Also, nonlinear MPC (NMPC) is employed using Hammerstein-Wiener model, which is either linearized or nonlinear optimization problem is solved using interior point method. Proposed workflows/frameworks are tested using a field-scale model of a SAGD reservoir located in northern Alberta, Canada. Both “scenario reduction” sampling method and PCE metamodel significantly reduce the computational cost while obtaining reasonable posterior distribution and production forecast. In addition, all four MPC variants successfully control the subcool in real-time, leading to lower cumulative steam-oil-ratio (cSOR) and more than 20% increment in the net present value (NPV). Practical implications of the proposed research will be consequential in designing accurate and energy efficient CLRM workflows while satisfying the constraints offered by the SAGD surface facilities, reducing carbon footprints, and improving economics.

• Subjects / Keywords

  • Real-Time Production Optimization
  • Model Predictive Control
  • Assisted History Matching
  • Closed-Loop Reservoir Mangement
  • Ensemble Kalman Filter

• Graduation date

  Spring 2018

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/R33F4M34H

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.