Using Geostatistical Inversion of Seismic and Borehole Data to Generate Reservoir Models for Flow Simulations of Magnolia Field, Deepwater Gulf of Mexico

Abstract

3Fugro-Jason Summary Study of key parameters of reservoir viz, porosity, water saturation, permeability and pore size distribution from well logging data is more complicated in carbonate reservoir due to geological heterogeneities than Clastic reservoir. The Magnolia field is located in GOM blocks GB 783 and 784 and produces from Plio-Pleistocene turbiditic sands that form a complex channel/levee sequence penetrated by 16 boreholes. The primary pays consist of two sands, each about 200 feet thick, separated by a 15 foot shale layer. The pays are divided into an eastern gas prone province and a western oil prone province. A reservoir flow simulation model is planned to optimize production from existing wells and to facilitate future field development. Construction of an accurate model is complicated by MDT pressure measurements which indicated compartmentalization below the resolution of conventional seismic analysis, and by overlap of the seismic attributes derived from producing reservoirs, wet sands, and shales. To mitigate these factors, geostatistical inversion was chosen to produce the rock property inputs for the flow simulation models. This approach allowed development of a rock properties model consistent with core data, log data, and geologic constraints as well as seismic information. It also allowed assessment of uncertainty through the generation of a statistically significant number of internally consistent alternate solutions (realizations). A Markov Chain Monte Carlo method was employed to integrate borehole and geologic information to produce acoustic impedance and lithology volumes which were then used to co-simulate porosity, permeability, p-wave velocity, and water saturation volumes. Multiple realizations of these products were reviewed, uncertainty was assessed, and a rock properties model was selected for conversion to a flow simulation modeling format. The entire process can be rerun relatively quickly to accommodate additional wells and improved seismic data or to match production history.