Data Mining and Optimization in Steam-assisted gravity drainage process

• Author / Creator

  Li, Chaoqun

• Steam-assisted gravity drainage (SAGD) is an enhanced oil recovery (EOR) technology widely used in Canada. Data available in SAGD industrial processes contain valuable information for monitoring, soft sensing, control, and optimization. This thesis focuses on data mining and optimization in the SAGD process, including the comparative study of machine learning algorithms, Bayesian Optimization and soft sensor design. Subcool is a key variable of SAGD, that is important for safety and efficiency. The popular machine learning algorithms, especially those having good practical merit, are tested and compared in a subcool monitoring case. The advantages of multiple machine learning algorithms are analyzed and discussed. Moreover, investigating the original dataset suggests that data quality and priori knowledge play a vital role in applying machine learning to study data analytics problems to oil sands industry. Bayesian Optimization considers model building and optimization simultaneously. Locally weighted quadratic regression is incorporated into the Bayesian Optimization framework, serving as the surrogate model. Details of the existing framework are tailored to use locally weighted quadratic regression as the surrogate model. Numerical cases are tested to demonstrate the usefulness of the Locally Weighted Quadratic Regression based Bayesian Optimization (LWQRBO). Additionally, the application of the proposed LWQRBO to address the formulated SAGD optimization problem is explored. Finally, optimization results show the applicability of LWQRBO in the SAGD process. Soft sensing constructs models with data mining approaches. It is an alternative way to measure process variables when hardware sensors are not available. Stacking online soft sensor is designed for the SAGD process. The idea of stacking models inspires this design which applies multiple linear regression as a second level model, considering ease of implementation. Two cases are studied to show the applicability and suitability of the designed soft sensor in the SAGD process. The topics of Bayesian Optimization and ensemble models for soft sensor extend promising research directions. They are presented at the end of this thesis.

• Subjects / Keywords

  • Data Mining
  • Optimization
  • SAGD

• Graduation date

  Spring 2018

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R33N20W0R

• 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.