Generating a Regression Model Proxy for CO2 storage

Abstract

CO2 storage is regarded an important asset in reducing total CO2 emissions to the atmosphere. Several methods for storing CO2 have been proposed, but underground storage in saline aquifers are among the most promising. Storing CO2 underground is a comprehensive process that requires thorough understanding of aquifer behavior which is acquired through reservoir simulations, which are time consuming and data demanding. Injection is an expensive process and to save cost it is desirable to optimize the injection process. Optimization of injection scenarios require many reservoir simulations. It is desirable to save time on simulating different injection scenarios, and proxy can be created to take over for the simulator. In this thesis a regression model proxy is being built to replace the need for reservoir simulations and to help optimize the injection scenario.Creating a proxy requires a thorough understanding of the injection process and many simulations has to be conducted. To reduce the amount of simulations required the input parameters can be scaled dimensionless. Still there are many simulations required to generate enough data for the proxy to use. The process can be simplified by writing computer scripts to automate simulations and generation of the proxy. Results prove that it is possible to create a regression model proxy for CO2 injection scenarios and to use it to find an optimal injection scenario.