Vida Gholami

Date of Graduation

2014

Document Type

Dissertation

Degree Type

PhD

College

Statler College of Engineering and Mineral Resources

Department

Petroleum and Natural Gas Engineering

Committee Chair

Shahab D Mohaghegh

Committee Co-Chair

Sam Ameri

Committee Member

Ebrahim Fathi

Committee Member

Jitendra Kikani

Committee Member

Marcello Napolitano

Abstract

CO2-EOR projects are becoming increasingly popular. Since enhanced recovery processes are applied to the mature fields, it usually involves a large number of wells. While the large number of wells leads to a better geological model, it results in very large flow models that are hard to manage, history match, and use as an optimization base. Nevertheless, injection-production optimization remains the core of all modeling efforts in CO2-EOR projects.;The objective of this work is to investigate the feasibility of using state-of-the-art data-driven proxy models to facilitate injection-production optimization in a CO2-EOR process. The use of coupled grid-based---SRM G and well-based---SRMW Surrogate Reservoir Model (as a proxy that runs in seconds) will be investigated as a tool to achieve the objectives of this study. The coupled SRM is built based on a reservoir simulation model that is developed for this purpose. The coupled SRM will be able to identify the dynamic reservoir properties (pressure, saturations, and component mole fraction at gridblock level) throughout the reservoir, along with the production characteristics at each well. It can be used to identify the optimum injection strategy (volume, rate, etc.) that would result in increased oil production.;The EOR technique that is attracting the most new market interest is CO2-EOR. First tried in 1972 in Scurry County, Texas, CO2 injection has been used successfully throughout the Permian Basin of West Texas and eastern New Mexico. The SACROC field, a depleted oil field located in western Scurry County in Texas, is the subject of this study.;A high resolution geological model was built for the northern platform. The model is based on a comprehensive geological study including 3D seismic survey and well logs. The porosity and permeability data for the fine grids were obtained from the Bureau of Economic Geology (BEG). The very long run-time of the reservoir simulation model that is the result of complexity of the reservoir makes it impractical to perform any sensitivity analysis, uncertainty analysis, or optimization study on the model. In order to overcome this problem, developing a surrogate reservoir model based on Artificial Intelligence and Data Mining techniques was planned. The coupled SRM provides the means for performing a large number of simulation runs, in short period of time, to be used for uncertainty quantification, and search of solution space for optimization.;Multiple injection scenarios were designed and run using a numerical reservoir simulator. The results were used in order to build a comprehensive spatio-temporal dataset, which includes all aspects of the reservoir model that is needed to train, calibrate, and validate the coupled SRM. From the parameters assimilated to form the comprehensive spatio-temporal dataset, Key Performance Indicators were identified and ranked. These KPIs helped to determine the dimensionality of the input space used to develop the SRMs (SRM W and SRMG).;Optimization may be identified by two focus areas. Building an efficient evaluation function and finding the quickest path to global minima. In this work, we focus on the efficiency of the evaluation function. The integrated SRM was built by coupling the two aforementioned SRMs. This SRM can be used to identify the optimal injection strategy (volume, rate, etc.) that would result in increased oil production while keeping an eye on the flood front.

Download

Share

COinS