Quantify reliability of storage capacities estimates

Quantify reliability of storage capacities estimates

State of the art

Dynamic capacity assessment relies strongly on modelling capabilities. Schlumberger ECLIPSE has been a benchmark reservoir simulation tool for the past 25 years to deal with increased complexities in oil fields, as well as alternative hydrocarbons and CO₂ geological storage.

Then next-generation reservoir models (e.g. INTERSECT, simulation code developed by Schlumberger), which will also be applied to study injectivity, include scalable parallel processing capability for dramatic decrease in run-times on multi-million segment grids, advanced gridding techniques for accurate representation of heterogeneities (DeBaun et al., 2005; Dogru et al., 2009). INTERSECT was used in a recent case study, for the Australian Gorgon Project, taking advantage of its high performance with parallel processing (Edwards et al., 2012). However, these capabilities have not been used to assess the impact of heterogeneities such as permeable faults and (low permeability) granulation seams on plume migration and ‘bankable’ storage capacity estimates in faulted and complex reservoirs targeted for CO₂ storage.

Within the flow modelling workflow for dynamic capacity estimates, uncertainties are introduced from a range of sources, for example uncertainties in geological characteristics and limitations of modeling input and software. Uncertainty can take several forms. A first category referred to as parametric stems from difficulty in estimating the input parameters (in a broad sense) of models/analysis due to the limited number, poor representativeness (caused by time, space and financial limitations), and imprecision of observations/data. Examples are model spatially homogeneous parameters like multiphase flow properties of a given rock formation and geological heterogeneities related to spatial variability (e.g., heterogeneous reservoir permeability field). A second category referred to as modelling uncertainty is related to the assumptions underlying the construction of the model.

Examples in the domain of CO₂ storage modeling are the inclusion of the capillary processes, the size of the grid mesh and the choices of boundary conditions. This last category also includes modelling errors, which can be defined by the discrepancy between the simulation code and the actual physical system (e.g., Kennedy & O'Hagan, 2001).

Any efficient management of storage site (e.g., site selection, injection operations, etc.) should then rely on a “good picture” of what is unknown: this is the purpose of Uncertainty Quantification (UQ). Standards for reliability of reserves in the domain of O&G typically rely on quantiles P10, P50, P90 to do so (Px is a statistical confidence level for an estimate, when probabilistic Monte Carlo type evaluations are adopted. Px is defined as x% of estimates exceed the Px estimate. P90 and P10 are low and high estimates respectively).

Progress beyond the state of the art

INTERSECT will be used to investigate the impact of fracture networks and granulation seams on capacity evaluation.
Detailed analysis of these features, using down-hole characterisation data from the GeoEnergy Test Bed. Also, followon activities from the INTERSECT simulations of injectivity (Hontomin test site) will investigate the role of fractures on capacity evaluation in fractured carbonate reservoirs. These studies will improve our understanding of storage capacity estimates in complex reservoirs.

Outcomes

• Modelling techniques for high resolution simulation of fractured reservoirs and an understanding of the effect of fractures on storage estimates
• Methods and a modelling strategy (framework) for tackling the different uncertainty sources (parametric, modelling, spatial uncertainty, and model errors) for capacity estimates. P10, P50, P90 estimates (similar to the widely used reserve estimation techniques in the O&G) of the expected capacities of at least two sites of ENOS project fields.