Engergy and Mineral Engineering

Open Access
Author:
Lai, I-AN
Graduate Program:
Petroleum and Mineral Engineering
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
June 13, 2016
Committee Members:
  • Turgay Ertekin, Thesis Advisor
  • John Yilin Wang, Committee Member
  • Eugene Morgan, Committee Member
Keywords:
  • Artificial Neural Netwrok
  • Waterflooding
Abstract:
Water flooding is a predominant secondary recovery method used in conventional petroleum reservoirs. The performance of a water flooding project will be impeded if a free gas phase arises in the reservoir. The main purpose of this study is to apply artificial neural network technology to solve problems related to water flooding projects in three-phase reservoirs. An artificial neural network (ANN) is a computational tool developed based on biological neural systems. The computation of an artificial neural network is similar to the way the human brain can predict the results based on previous knowledge gained through experiencing various situations. This study presents three artificial neural network proxy models that are applied to predict production profiles, project design parameters, and reservoir properties. Users can save time by using these ANN models instead of using numerical simulations and thus can achieve more desirable recovery targets of water flooding projects. In this thesis, various scenarios are created by the combination of reservoir properties, project design parameters, PVT parameters, and rock and fluid parameters. Then, a commercial reservoir simulator is utilized to generate the production profiles of oil, water, and gas in water flooding projects. An artificial neural network is developed with the ascertained properties (reservoir properties, project design parameters, PVT parameters, and rock and fluid parameters) as inputs with production rates and time intervals as outputs. In the final stage, graphical user interfaces are created to facilitate users’ access to these ANN models. Users can input the parameters and the results are displayed in the graphical user interface. The selected ranges of input data are also displayed at the end through a graphical user interface in order to enhance its usability.