Date of Graduation

2018

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Petroleum and Natural Gas Engineering

Committee Chair

Ilkin Bilgesu

Committee Co-Chair

Samuel Ameri

Committee Member

Kashy Aminian

Committee Member

Ilkin Bilgesu

Abstract

The Marcellus shale is one of the largest unconventional gas shale plays in the United States. It underlies much of Pennsylvania, West Virginia, and New York and even extends under Lake Erie and into Canada. The most effective methods to produce from this play is to drill horizontally into the shale formation and use hydraulic fracturing. Hydraulic fracturing creates pathways for the hydrocarbon to flow from the shale into the wellbore. These lateral sections of the well are typically completed from toe to heel over a number of stages using a plug and perforate method.;This research focuses on a Marcellus shale well drilled in Morgantown, West Virginia. The well was completed using five different fracture designs over a total of 28 stages throughout the lateral. This well served as more of a learning experience than a typical horizontal well. A flow scanner was also run through the well after hydraulic fracturing to discover more information that is typically not acquired in most wells. All data for this well was provided by the Marcellus Shale Energy and Environment Laboratory (MSEEL) research group. The MSEEL participants were Northeastern Natural Energy, Department of Energy, West Virginia University, Ohio State University, and others who reviewed all the collected information. The goal of research group was to improve the understanding the shale characteristics in this region in order to be more efficient in the completion of other wells in this location.;A neural network model was used to examine the efficiency and performance of different completion methods and their impact on gas production. Several input parameters such as plug depth, total shots, natural fractures, measured slurry, measured clean fluid, measured proppant, pump time and stage length were used to predict gas flowrate. Several combinations of training, validation, and testing sets were employed with different number of hidden layer neurons and the best combination was determined.

Download

Share

COinS