Open Access
Shihab, Rizvi
Graduate Program:
Petroleum and Natural Gas Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
April 25, 2010
Committee Members:
  • Turgay Ertekin, Thesis Advisor
  • Robert W Watson, Thesis Advisor
  • Li Li, Thesis Advisor
  • Larry Grayson, Thesis Advisor
  • in-situ combustion neural networks
The main topic of the research is the enhanced oil recovery (EOR) method of forward dry in-situ combustion (ISC). ISC is an EOR method used to produce heavy oils with high viscosity levels that are either infeasible or not economical using other EOR methods. The ISC process is a thermal recovery method that initiates when hot air is injected into the injection well to create heat inside the reservoir, which will in turn create a burning front moving from the injection well toward the production well. During this process, some of the reservoir's oil will be utilized as fuel by the process of in situ oil burning. The fuel that is produced by this process will support the combustion front. Once the heat is generated steadily by the continuous injection of hot air, the oil viscosity in the reservoir will be decreased which allows the new less viscous oil to flow to the producer. The project's goal is to create an expert system for forward in-situ combustion that has the ability to predict similar outcomes to those obtained by a thermal recovery simulator . The predicted outcomes are the oil production, the gas production and the abandonment time of the project. In order to develop the expert system, relevant output results of oil production, gas production and abandonment time of the project need to be obtained using a thermal simulator for three field patterns with varying sizes ranging from five acres to 25 acres. The numerical simulator uses ten input variables including field properties and design parameters. Such examples of these inputs are porosity, permeability, injection rate, oxygen content of injection, thickness of the reservoir, initial temperature and pressure of the reservoir and the initial oil and water saturation. Due to these different variables, it is expected that the output from the simulations will have a large range and wide scope, which will help in developing a useful and flexible expert system. The simulations generated data for three different sized patterns ranging from five to twenty five acres. The next objective in the project was to create an expert system for each different pattern using artificial neural networks (ANN) . ANN is a tool that functions like the human brain. It is a mathematical model that uses an inter-connected set of neurons that is able to adapt itself depending on the data fed to the system. Because of the characteristic of continuous adjustment, ANN is called an adaptive and a non-linear system. The ANN, when modeled correctly to fit a specified set of data, will be able to spot trends in the data when going through the learning process. If the learning process is successful, the ANN system will be able to predict the simulated data within a certain level of accuracy. The targeted level of accuracy for this work was five percent error or lower. The process meeting this target is called the validation process and in this project the average error of the ANN system is found to be below five percent, which deemed the expert system to be successful.