A case study of dynamic modeling of multiple regionally-extensive reservoirs using a unified fluid description

Abstract

A novel approach to perform regional scale simulations for a group of Gulf of Mexico reservoirs has been performed. The main objective of such an effort has been to integrate reservoirs with a potential dynamic communication through a single sensible initialization model in the simulator. In the past, some of these reservoirs were modeled standalone, and the current information has forced us to connect multiple reservoirs that are connected laterally. Therefore, performance of multiple Gulf of Mexico (GOM) reservoirs was assessed using region-wide dynamic simulation many reservoir simulation studies, at the initial phase of this project, “isolated” performance prediction of the subject reservoirs was considered, mainly due to unknown lateral connectivity and the differences in the discovery dates. In later years, there was a need to predict the unified system performance based on surveillance data, and to be able to test the possibility of the connectivity of the multiple reservoirs and reservoir segments. The reservoir fluid system analysis becomes even more challenging when we have a system with vertical and areal compositional gradients such as the system presented here. In addition to the surveillance production data, available data also included basin wide geochemical analysis that addressed reservoir compartmentalization. Such data confirmed the well responses and assisted in describing and initializing the reservoir correctly. Along with the analysis of the particular case to be discussed here, the reasons behind the graded oil/gas reservoirs will also be discussed and some of the critique will be provided to highlight the broader implications of field observations and as well as some of the ideas presented in the literature. Furthermore, we have integrated both standard black oil PVT (pressure–volume–temperature) and surveillance data (including geochemical information). Such effects proved that the overall performance can be represented better using a single extensive model as supported by the history matched model. Some of the issues related to non-equilibrium gradients were also highlighted.