Optimization of chemical enhanced oil recovery methods for naturally fractured carbonate reservoirs

Mejia, Miguel, M.S. in Engineering
Journal Title
Journal ISSN
Volume Title

Carbonate oil reservoirs are important energy sources, accounting for over 60% of the world’s oil reserves. Recovering oil from these reservoirs is challenging, especially if they are naturally fractured. Waterflooding is inefficient because water flows through the highly permeable fractures and bypasses the rock matrix, where most of the oil is stored. Mixed-wettability, low matrix permeability, and large heterogeneities also make secondary oil recovery challenging. Chemical enhanced oil recovery with alkali, surfactants, and polymer addresses some of these challenges. Surfactants can lower the interfacial tension to decrease the residual oil saturation. Polymer increases the viscosity of the injected water, improving the microscopic and macroscopic sweep efficiency. This research involves the optimization of some chemical flooding methods for naturally fractured carbonates. Coreflood experiments and the UTCHEM reservoir simulator were used to investigate alkali-surfactant flooding in fractured Texas Cream limestone cores. A decrease in fracture mobility caused by viscous phase trapping in the fracture was identified as the main reason for the high observed oil recoveries. Due to the uncertain properties of the viscous phase trapped in the fracture, polyethylene oxide (PEO) polymer was investigated for mobility control. Coreflood experiments demonstrated the viability for using PEO in 18 mD cores. PEO significantly improved oil recovery in a fractured core. The viscosity and cloud point of the PEO were systematically investigated. The polymer concentration, temperature, salinity and hardness were varied, and several additives were added to potentially increase the range of conditions for which PEO could be applied to EOR. Methyl-urea, urea, and ethanol were identified as additives to increase the cloud point and viscosity of PEO. Finally, machine learning models including support vector machine, random forest, and neural network models were trained to predict the aqueous stability of surfactant solutions and phase behavior of microemulsions. A large database of over 600 phase behavior experiments and over 800 aqueous stability experiments was used to train the models. The models may be used to guide the process of selection of surfactants that produce sufficiently high solubilization ratios.