Object Characterisation and Simulation of Thermal Recovery from Karstified, Brecciated and Fractured Bitumen Carbonate Reservoirs

Abstract

Abstract A continued increase in energy demand has amplified the significance of commercial heavy oil and bitumen recovery from complex carbonates formations such as the Grosmont Formation (OOIP ~ 406.5 billion barrels) in Alberta, Canada. To facilitate commercial development of bitumen carbonates, we have designed reservoir simulation models of complex carbonate reservoirs based on the concept of multiple interacting objects. Spatial distribution of different objects including fractures, vugs, breccia, and matrix are constructed by using stochastic methods with intensity functions derived from cores, logs, drilling and geologic data. Thermal reservoir simulations are conducted directly on realizations of these 'objects network' reservoir models. Although data from the highly fractured, karstified and vuggy bitumen-rich Grosmont Formation is used in this paper, this methodology is generic and applicable to other complex carbonate reservoirs. Results suggest that continuous type steam-based enhanced oil recovery (EOR) such as steam-assisted gravity drainage (SAGD) may not be best suited for bitumen recovery from complex carbonates. Introduction The ultimate aim of reservoir characterization is to construct a representative spatial quantification of storativity (porosity), hydraulic conductivity (permeability), and fluid phase saturations. In highly complex carbonates where fractures, vugs, matrices, and karsts contribute to recovery performances, reservoir models must sufficiently represent the heterogeneity in hydraulic corridors (described here as object clusters) to accurately predict fluid breakthrough and ultimate recovery for different EOR technologies. Unfortunately, the 'forward modeling' approach (which focuses on understanding drivers that generated fractures by analyzing parameters such as stress distribution, fracture height, fracture spacing) often used by geoscientists for characterizing naturally fractured reservoirs (NFR) of sandstone matrix is seldom sufficient for carbonates. This is primarily due to the complex process of diagenesis inherent in carbonates. As a result, a systematic combination of the 'forward modeling' approach to the 'inverse modeling' approach (this approach focuses on understanding the responses created by fractures such as productivity heterogeneity, breakthrough, and channelized flow) is favored for the Grosmont carbonate reservoir. There are seldom sufficient data for complex carbonates, especially because of the difficulty to obtain consolidated sample representative of the tremendous heterogeneity. Although the emergence of tools such as the Formation Microimager (FMI), Computed Tomography (CT) scans, Scanning Electron Microscopy (SEM) and the improvements in traditional formation evaluation methods have contributed to increasing data availability, effective integration of data at different scales is extremely important to derive value from these measurements. Although statistics derived from wellbore (typically from vertical wells) measurements provide insight into the vertical distribution of properties such as fracture geometry, fracture length, fracture orientation, vugs, karsts; spatial distribution of these properties can be constrained by the knowledge of larger (km) scale correlations. As an example, previous studies suggest that a good large (km) scale lateral continuity of facies exist for the Grosmont Formation (Edmunds et al., 2009). Geologic studies have also described the predominant location of large karsts (nearer the sub cretaceous unconformity), (Hans et al., 2012). Therefore, in addition to well data, larger scale seismic and geologic data offer increased data control points thereby reducing the uncertainty in the developed model. Analysis of early pilot tests (Ezeuko et al., 2013) indicates a reasonable-to-high injectivity, suggesting a high degree of communication between high conductivity (mostly fractures, vugs, and karst) object clusters.