Data-Driven Proxy Models for Assisted History Matching of SAGD Reservoirs

• Author / Creator

  Jain, Tarang

• In reservoir simulation studies, history matching is extensively used for uncertainty reduction and reservoir management. History matching using Ensemble Kalman Filter (EnKF) is a promising approach due to its non-iterative nature and ability to assimilate a large number of model parameters. However, in processing a large set of realizations, this method suffers from high computational time and cost associated with the use of commercial reservoir simulators. Therefore, there is a scope for some improvement in this approach especially in the case of complex thermal recovery process such as steam assisted gravity drainage (SAGD). In this work, the computational cost is reduced significantly by developing proxy models that can substitute the need of reservoir simulator during the assisted history matching process. Different proxy models such as Polynomial Chaos Expansion (PCE) and Artificial Neural Networks (ANN) are tested to represent the outputs of the conventional reservoir simulator. Permeability realizations of the SAGD reservoir are first parameterized using Karhunen- Loeve (KL) series expansion and represented in the form of uncorrelated random variables. The developed proxy models utilize random variables obtained from KL expansion as input parameters. Proxy models are further integrated with EnKF algorithm as a substitute for reservoir simulator. Computational requirement of the proxy model during the development as well as deployment as compared to commercial reservoir simulator is emphasized in this study. The proposed approach is validated using a field-scale SAGD case study of northern Alberta. The observed daily oil rate, cumulative oil production, and cumulative steam to oil ratio are history matched using the proposed method. Results show that as compared to conventional EnKF, the integration of data-driven proxy models can perform assisted history matching in quick, low-cost manner while maintaining the accuracy of results. This work has a potential to cut down the monetary and time constraints during the assisted history matching process.

• Subjects / Keywords

  • Artificial Neural Networks
  • Polynomial Chaos Expansion
  • Karhunen Loeve Expansion
  • Proxy Models
  • History Matching
  • SAGD Reservoirs
  • Ensemble Kalman Filter

• Graduation date

  Spring 2017

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R31V5BS04

• 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.