Probabilistic approach to CO2 plume mapping for prospective storage sites: The CarbonNet experience.

摘要

In CO2 storage, there is a requirement to predict the range of possible plume extents and travel paths and associate a probability with this range, given inevitable subsurface uncertainty. The probabilistic expectation of the plume at future times is used for assurance that the plume will remain within the defined storage boundaries (both geographical and stratigraphie) for the required time with a high level of confidence. Australian GHG storage regulations call for a prediction of all plume paths with more than 10% probability of occurrence (i.e. plume paths at P90 confidence level). Here we outline a fully three-dimensional probabilistic approach based on reservoir modelling sensitivity and uncertainty analysis, adapted from the petroleum industry and suitable for high-mobility CO2 plumes in thick and well-defined reservoirs. The method can also be extended to other basins and geological circumstances. In the petroleum industry, i is commonplace to evaluate resources in probabilistic terms with some one-dimensional objective parameter such as oil in place, recoverable reserves, or nett present value. This methodology can be adapted easily to objective measures such as vertical ascent of a plume relative to a caprock or lateral approach of the plume to a boundary or other geographic feature to be avoided (e.g. a mapped fault). What is novel in our approach is to analyse plume paths (extents) in a fully three-dimensional statistical manner to generate probabilistic maps and cross-sections of plume extents probability in comparison to physical features such as vertical or lateral boundaries to inform on containment risks and areas with key monitoring requirements. Maps of 90% probability of the plume path from the actual dynamic simulation runs and cross-sections of the 3D plume probability are illustrated to meet Australian regulatory requirements, and to optimizeselection of MMVmethods and locations. The CarbonNet Project aims to store a nominal 125 million tonnes of CO2 over 25 years in the same basin still in use for hydrocarbon extraction, and adjacent to an important onshore aquifer. Plume management and containment is therefore vital and high confidence must be placed on plume path modelling, including the analysis of rare statistical outliers. The new approach described here offers a method to quantify plume path uncertainty in terms of lateral extent (i.e. map view) and vertical extent (cross-section), offering 3D understanding of plume containment with an appropriate high level of regulatory and public confidence.