Date of Award

January 2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Petroleum Engineering

First Advisor

Adesina S. Fadairo

Abstract

This study investigates improved strategies for enhancing oil recovery (EOR) in challenging reservoir conditions by combining chemical surfactant flooding and data-driven optimization techniques. Given rising global energy demand and the depletion of conventional oil reserves, chemical EOR methods, particularly surfactant flooding, are gaining popularity for their ability to mobilize residual oil via mechanisms such as reducing interfacial tension, altering wettability, and inhibiting asphaltene formation. The research begins by examining the effectiveness of surfactant applications across various reservoir conditions, emphasizing the need for surfactants capable of withstanding challenging environments characterized by high temperatures, high salinity, and low permeability. Laboratory experiments demonstrated that zwitterionic surfactants significantly improved wettability in Middle Bakken core samples, reducing interfacial tension from 34.5 mN/m to as low as 8.9 × 10⁻¹ mN/m, resulting in an additional oil recovery of 16.4% during spontaneous imbibition tests. A spectrophotometric approach was then utilized to investigate the adsorption behavior of zwitterionic surfactants on Bakken minerals. Modeling adsorption with Langmuir isotherms revealed complex interactions influenced by salinity, temperature, and surfactant functional groups. To complement the experimental work, a hybrid approach was developed by combining a robust numerical reservoir model with data-driven predictive modeling. A detailed reservoir model of the Middle Bakken Member, incorporating multistage hydraulic fractures, was used to simulate surfactant Huff-n-Puff processes using CMG-STARS. This model captured the effects of parameters such as surfactant adsorption, concentration, and injection dynamics, providing a realistic representation of reservoir behavior. Concurrently, a Gradient Boosting Machine (XGBoost) was employed alongside Response Surface Methodology (RSM) for predictive modeling and optimization, achieving high predictive accuracy and enabling systematic sensitivity analysis of key operational parameters. The integration of numerical simulations with machine learning revealed a good correlation between actual and predicted oil recovery factors, underscoring the critical role of surfactant concentration and adsorption in optimizing recovery outcomes. This comprehensive methodology bridges traditional reservoir simulations with modern machine learning approaches, offering a flexible, accurate, and cost-effective strategy to optimize EOR processes. By exploring parameter sensitivities and optimizing surfactant injection strategies, this study provides valuable insights for enhancing operational efficiency and extending reservoir life, ultimately demonstrating the potential for advanced data-driven optimization to drive innovation in the petroleum industry.

Download

Share

COinS