An Investigation of the Efficacy of Advanced Well Structures in Unconventional Multi-phase Reservoirs

Open Access
Enyioha, Chukwuka
Graduate Program:
Petroleum and Natural Gas Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
January 08, 2015
Committee Members:
  • Turgay Ertekin, Dissertation Advisor
  • Zuleima T Karpyn, Committee Member
  • Yilin Wang, Committee Member
  • Daniel Kifer, Committee Member
  • Unconventional
  • advanced well structures
  • artificial intelligence
  • production prediction
  • field development
  • history matching
Advanced well structures have continued to see an increase in use and field application, particularly for unconventional reservoirs. Advances in such fields as geo-steering, directional drilling, and measurement while drilling, have enabled the possibility of constructing wells with different kinds of configuration and extended reach. These wells are constructed to increase contact with the reservoir, thus enhancing productivity whilst reducing surface footprint by the ability to access isolated productive zones. The complexity of these well structures makes it an arduous task to develop analytical models to predict performance. Traditional methods for well performance analyses barely capture the complex interactions between these well structures and the reservoir. In this study, artificial expert systems using artificial neural networks are designed and developed for various tasks involving advanced well structures in unconventional multi-phase reservoirs. Artificial expert systems have the ability to learn information that exists between a pair of input and output without necessarily being encumbered by the need for knowledge of the physics of the relationship between the pair. These systems establish functional correlations betwen the input and output of any system that can subsequently be used to estimate the response of the system, given a new input. Three sets of artificial expert systems have been developed in this study that reflect the conventional needs of the industry. The Production Performance Expert System (PPEx) is developed to predict the performance of advanced well structures in tight oil reservoirs given that some knowledge of the reservoir properties is available. The Well Design Expert System (WDEx) is developed to predict a possible advanced well structure that can meet a desired production target given that some knowledge of the reservoir properties is available. This can be seen as working in the inverse direction of the PPEx system. The WDEx system is particularly useful in the field development planning stage of any project. The History Matching Expert System (HMEx) is developed to predict possible values of reservoir properties given an advanced well structure and its production profiles. This can also be seen as working in the inverse direction of the PPEx system. The HMEx system is important for better characterization of the reservoir. These three sets of expert systems are ultimately combined in such a fashion that makes each one complementary to the other. Once combined, this tool becomes a useful and relatively inexpensive means for quick and efficient analyses of advanced well structures available to the reservoir engineer. The developed models show good performances as seen in results of various tests applied. Generalization test results for the PPEx systems show average prediction error of 8%. Similar results are obtained for other expert systems. These results lend credence to the application of artificial intelligence for such complex wellbore and reservoir systems.