--> Reservoir Modelling Insights From Experimental Stratigraphic Analogs

AAPG Annual Convention and Exhibition

Datapages, Inc.Print this page

Reservoir Modelling Insights From Experimental Stratigraphic Analogs

Abstract

Stratigraphic architecture and internal heterogeneity are important parameters for measuring and modelling reservoir performance, but how much complexity should be used to model a reservoir? Numerical forward models are generally scalable, although their properties are typically assigned deterministically or are populated stochastically. Natural stratigraphic analogs contain complexity at multiple length scales, but they tend to be difficult to relate to reservoir-scale observations. Experimental stratigraphic analogs offer useful insights for reservoir modelling because they possess a wide range of variability that are similar to natural systems, and they can be sampled at multiple scales. Fluvial depositional systems are complicated because they are spatially and temporally discontinuous. For these reasons, we use an experimental model of multiple-interacting braided streams to examine influences of spatial heterogeneity on reservoir performance, and to quantify upscaling in fluvial environments. The experimental basin contained distinct depositional facies associations associated with four margin-sourced distributive fluvial systems (DFS) and an axial river that recorded deposition during six sediment-flux cycles and two subsidence regimes. The dataset was sampled at different resolutions to generate synthetic wells and seismic, and to populate properties from a common data source. Stacking patterns, property heterogeneity, and architecture were used to understand the impact of complexity on flow responses at different levels of captured complexity. This approach permits testing of three-dimensional sampling routines that characterize and quantify vertical and lateral complexity. Our conclusions have broad implications for reservoir modelling because they highlight the balance between capturing complexity and the costs of sampling, and they can be applied to systems where three-dimensional heterogeneity can only be measured using sparse datasets.