Loading...

Development of Low-order Model/controllers for Oil Reservoir Smart Wells

Hemmati, Sahar | 2021

375 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54474 (06)
  4. University: Sharif University of Technology
  5. Department: Chemical and Petroleum Engineering
  6. Advisor(s): Pishvaie, Mahmoud Reza
  7. Abstract:
  8. Upstream oil industry operators have turned to optimal production methods for further recoveries, such as smart wells. Control algorithms for smart wells must be equipped with online measurements, in other words, reflected in the feedback philosophy. On the other hand, after discretization, an oil reservoir's dynamic and control-driven model becomes a realization or a large-scale and sparse state space. One way to deal with this problem and provide an appropriate model for the design of smart well controllers is to use appropriate reservoir dimensional reduction methods. Therefore, to design low-order controllers, a study must first be performed at the level of the reservoir model and its reduction. The purpose of this dissertation is to investigate and evaluate the comparative methods of model dimensionality reduction of large-scale systems of oil reservoirs to lower orders and appropriate models in both data-driven and model-oriented approaches. The studied techniques and approaches rely mainly on system theory and engineering research algorithms for online and real-time control, history matching, and optimization. Two models of single-phase one-dimensional reservoir and two-dimensional two-phase synthetic reservoir have been selected to implement these approaches. Each of these systems is modeled by Balanced Truncation (BT) and Proper Orthogonal Decomposition (POD), which are model-driven, and Dynamic Mode Decomposition (DMD), and Dynamic Mode Decomposition with control (DMDc), which are data-driven, reduced and compared with the results of the least square and the Numeric algorithms for Subspace state-space system Identification (N4SID) methods. All methods can achieve the reduced state-space model of one-dimensional and one-phase and 5-spot two-dimensional and two-phase reservoir models. In the BT model-driven method, in the one-dimensional one-phase reservoir model, the order of the model is reduced from 51 to 6 and in the two-dimensional two-phase reservoir model from 242 to 14, It can be said that the BT method has better and faster results than BPOD in this research. The DMDc method also works well in identifying two-dimensional and two-phase reservoir models and can provide the preconditions for dimensional reduction. Therefore, the obtained results can be used to identify and control high-dimensional processes in real-time. These methods can model systems with many states and outputs online and provide a reduced model with a low computational cost by considering the minimum required dynamics to capture.



  9. Keywords:
  10. Proper Orthogonal Decomposition ; Balanced Realization ; Reduced-Order Controller ; Reduced-Order Oil Reservoir Model ; Balanced Realization ; Smart Wells

 Digital Object List

 Bookmark

No TOC