Machine learning approaches for enhanced analyses of rock properties in geologic CO2 storage

Barteneva, Ekaterina

Permalink

https://hdl.handle.net/2142/121385 Copy

Description

Title

Machine learning approaches for enhanced analyses of rock properties in geologic CO2 storage

Author(s)

Barteneva, Ekaterina

Issue Date

2023-07-21

Director of Research (if dissertation) or Advisor (if thesis)

Makhnenko, Roman

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Carbon storage
• machine learning
• caprock
• permeability
• breakthrough pressure
• wellbore logs
• acoustic emission.

Abstract

Geological carbon storage (GCS) has emerged as a promising approach for mitigating the accumulation of carbon dioxide (CO2) in the atmosphere. This research investigates the application of machine learning (ML) techniques to enhance the analysis and optimization of CO2 storage operations. The initial part of the study focuses on predicting permeability and breakthrough pressure in the caprock, the upper layer of GCS. By conducting a thorough literature review and developing robust ML algorithms, accurate predictions are achieved using input variables derived from the comprehensive datasets. Prediction of the CO2 breakthrough pressure is performed using three ML methods: the non-linear neural network, Bayesian framework, and the classification model are developed and evaluated for their accuracy and reliability. The analysis includes direct and indirect test results with two (porosity and permeability) and five (porosity, permeability, specific surface area, pore radius, and clay content) input parameter configurations. The results reveal strong correlations between the predicted and measured breakthrough pressure values. The non-linear neural network model demonstrates a superior performance when utilizing five input parameters, while the Bayesian framework yields identical predictions for both two-parameter and five-parameter configurations. The classification model successfully captures the correct range of values in its predictions for both configurations. The machine learning techniques are also utilized to analyze wellbore data and acoustic emission waveforms. Fractured wellbore intervals are predicted based on petrophysical and mineralogical parameters, underscoring the importance of considering multiple parameters for accurate fracture characterization. The analysis of acoustic emission waveforms demonstrates the effectiveness of deep learning models in denoising seismic data and accurately predicting the first arrival time. The findings of this study contribute to a better understanding of the correlation between the rock properties that are crucial for the analysis of the geologic CO2 storage. This work highlights the potential of ML approaches in robust assessment of caprock sealing capacity, enhancement of the quality of the seismic data interpretation, and efficient detection of the fractured intervals based on the wellbore log data.

Graduation Semester

2023-08

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/121385

Copyright and License Information

Copyright 2023 Ekaterina Barteneva

Owning Collections