Author ORCID Identifier

https://orcid.org/0009-0005-4293-3823

Semester

Summer

Date of Graduation

2024

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Petroleum and Natural Gas Engineering

Committee Chair

Shahab D. Mohaghegh

Committee Co-Chair

Samuel Ameri

Committee Member

Samuel Ameri

Committee Member

Kashy Aminian

Abstract

This study examines the application of artificial intelligence (AI) and supervised machine learning techniques to forecast production from unconventional shale wells, utilizing actual field measurement data over a period of two years. Traditional methods, such as decline curve analysis, offer valuable insights but often fail to fully capture the complex nuances affecting productivity and tend to rely excessively on empirical equations.

In this research, the AI-based Shale Analytics approach, introduced by Mohaghegh in 2017, is employed. This method leverages Big Data Analytics to identify unique patterns from actual field observations, enhancing the evaluation and quantification of various productivity factors, facilitating their comparison. The analysis encompasses over 400 shale wells in the Marcellus Shale, focusing on the impact of field measurements on well performance.

The TensorFlow and Keras machine learning libraries were utilized to develop an Artificial Neural Network (ANN) optimized for analyzing extensive datasets of field measurements. These datasets include variables such as well metrics, hydraulic fracturing details, formation properties, completion designs, and operational data. The ANN was developed to independently learn from this vast dataset, without reliance on previous models. The development process involved categorizing the datasets for training, calibration, and validation, enabling the ANN to accurately predict the production efficiency of existing wells. The model's predictive accuracy was further validated by applying it to new and analogous data sets that had not previously been encountered by the ANN.

The results demonstrate that production performance can be accurately predicted using a comprehensive dataset of actual field observations. The ANN model efficiently utilizes this dataset to identify optimal parameters for achieving high predictive accuracy. Additionally, the integration of AI models into reservoir management suggests potential enhancements in both the efficiency and accuracy of operational outcomes, highlighting areas where these technologies could improve industry practices.

Download

Share

COinS