Time-lapse gravity data at Prudhoe Bay: new understanding through integration with reservoir simulation models

Abstract

In this thesis, I revisit the time-lapse gravity data collected to monitor gas-cap water injection at Prudhoe Bay to understand how we can best integrate a field's detailed reservoir simulation data into the gravity inversion for an integrated interpretation. For this, I explore two methodologies that can directly incorporate into the inversions a set of time-lapse density models that can be constructed from the reservoir simulation parameters. The first approach is a continuous variable inversion where the reservoir information can be converted into an appropriate reference model to guide both the shape and amplitude of the recovered densities from the water flood. The second is a binary inversion, where the reservoir data are included as background density models to predefine the amplitude of the recovered time-lapse densities, allowing the algorithm to focus on recovering the locations and shape of the expanding water plume. Results for the two methods show that the reservoir data do guide and constrain the gravity inversions to distribution comparable to water movement within the simulation models, and thus provide valuable approaches for integrated monitoring with time-lapse gravity. The results additionally demonstrate deviations from the reservoir simulation models, indicating that either the surface gravity is not sensitive to the subtle changes in saturation in particular regions, or that the reservoir models contain saturation changes that may be too high or too low in areas.