Degree

Master of Science in Engineering (MSE)

Department

Petroleum Engineering

Document Type

Thesis

Abstract

Surfactant flooding is a promising chemical enhanced oil recovery (EOR) method for mobilizing residual oil through interfacial tension reduction and wettability alteration. However, its performance in associated-gas reservoirs can be limited by salinity, gas composition, surfactant adsorption, and possible chemical instability. This study investigates the effect of salinity on surfactant flooding efficiency using an integrated reservoir simulation and machine learning workflow. A synthetic three-dimensional reservoir model was developed in ECLIPSE using a 10 × 10 × 3 Cartesian grid to simulate surfactant flooding over a 300-day production period under salinity conditions ranging from 100 to 50,000 ppm. Key outputs, including field oil production rate (FOPR), block oil saturation (BOSAT), block total surfactant concentration (BTCNFSUR), and block total surfactant adsorption (BTADSUR), were analyzed to evaluate oil displacement, surfactant transport, adsorption, and reservoir conformance. The results show that salinity strongly controls surfactant flooding performance. Ultralow- and low-salinity cases produced the most favorable responses, with 400 ppm, 1000 ppm, and 15,000 ppm showing the best balance between oil mobilization and sustained late-time production. At 400 ppm, FOPR recovered from a late-time minimum of approximately 40–60 Sm³/day to a stable plateau of about 220–235 Sm³/day, while 1000 ppm and 15,000 ppm also maintained strong production support. In contrast, salinity above 25,000 ppm caused channelized flow, poor sweep efficiency, persistent high BOSAT, localized BTCNFSUR and BTADSUR, and late-time FOPR decline to approximately 50–60 Sm³/day, indicating severe conformance loss and reduced flood efficiency. Machine learning models were developed to predict production response from simulation-derived variables. Random Forest achieved the best performance, with R² = 0.9975, RMSE = 8.51, and MAE = 2.98. SHAP analysis identified BOSAT, salinity, and surfactant-related variables as dominant predictors. Overall, this study shows that salinity optimization is essential for effective surfactant flooding and that machine learning can support rapid screening and production forecasting.

Date

2-6-2026

DOI

https://proquest.com/docview/3347846086

First Committee Chair

Fathi Boukadi

First Committee Member

Nelson Chavez

Second Committee Member

William Chirdon

Download

Share

COinS