Exploring the effects of data quality, data worth, and redundancy of CO2 gas pressure and saturation data on reservoir characterization through PEST inversion

Abstract

This study examined the impacts of reservoir properties on carbon dioxide (CO2) migration after subsurface injection and evaluated the possibility of characterizing reservoir properties using CO2 monitoring data such as spatial–temporal distributions of gas pressure, which can be reasonably monitored in practice. The injection reservoir was assumed to be located 1,400–1,500 m below the ground surface such that CO2 remained in the supercritical state. The reservoir was assumed to contain layers with alternating conductive and resistive properties, which is analogous to actual geological formations such as the Mount Simon Sandstone unit. The CO2 injection simulation used a cylindrical grid setting in which the injection well was situated at the center of the domain, which extended out 8,000 m from the injection well. The CO2 migration was simulated using the latest version of the simulator, subsurface transport over multiple phases (the water–salt–CO2–energy module), developed by Pacific Northwest National Laboratory. A nonlinear parameter estimation and optimization modeling software package, Parameter ESTimation (PEST), is adopted for automated reservoir parameter estimation. The effects of data quality, data worth, and data redundancy were explored regarding the detectability of reservoir parameters using gas pressure monitoring data, by comparing PEST inversion results using data with different levels of noises, various numbers of monitoring wells and locations, and different data collection spacing and temporal sampling intervals. This study yielded insight into the use of CO2 monitoring data for reservoir characterization and how to design the monitoring system to optimize data worth and reduce data redundancy. The feasibility of using CO2 saturation data for improving reservoir characterization was also discussed.