ANALYSIS OF THE EFFICACY OF CARBON DIOXIDE SEQUESTRATION IN DEPLETED COALBED METHANE RESERVOIRS

Open Access
Author:
Liu, Liyang
Area of Honors:
Petroleum and Natural Gas Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
  • Turgay Ertekin, Thesis Supervisor
  • Turgay Ertekin, Honors Advisor
  • Luis Ayala H, Faculty Reader
Keywords:
  • coalbed methane
  • CO2 sequestration
  • Artificial Neural Network
Abstract:
In this study, the viability of Carbon Dioxide (CO2) sequestration in depleted Coalbed reservoirs is investigated using Computer Modeling Group LTD’s (CMG) Compositional & Unconventional Simulator (GEM). This simulator features dual-porosity and dual-permeability functions, and thus best suits the needs of the model intended. In order to imitate a stimulation fracture network around the horizontal well, a Stimulated Reservoir Volume (SRV) approach was implemented. Three different models with varied grid size, matrix properties, production rates, and injection rates were investigated in order to determine proper variable ranges for the Monte Carlo Simulation and the Artificial Neural Network (ANN) study, presented in the later part of the study. With low permeability and porosity, Coalbed methane cannot be easily produced, nor can CO2 be easily injected, without the implementation of fracture stimulation techniques. The SRV approach significantly improved case performances of both CH4 production and CO2 injection [1]. With varied production sand face pressure, production rates for each of the cases will be different. However, producers will be shut-in at a uniform minimum production rate of 300 MSCFD, followed by the opening of injectors at the same well location. Injection performances will be evaluated in this study. During the final stage of this study, three Artificial Neural Network tools were developed in order to predict various sets of data using combinations of input variables. The first tool can predict production and injection profiles of a given system with error very close or less than 20%. The second tool can predict wellbore design parameters and fracture characteristics with error less than 20%. The third tool can predict formation characteristics with error less than 20%, with the exception of one variable having larger error, yet within acceptable range.