About this webinar
Today’s reservoir challenges require a multi-disciplinary approach which integrates disparate pieces of information into the modeling of facies and reservoir properties. Good reservoir models must be sufficiently detailed to capture heterogeneity and flow characteristics and, at the same time, diverse enough to capture the uncertainties associated with our data and assumptions. This talk shows how such reservoir models can be produced within the framework of Bayesian geostatistics, and the benefits to Geoscientists, Engineers and Reservoir Modelers.
Facies modeling is particularly important since facies generally dominate the rock properties such as porosity and permeability, which, in turn, govern the overall static character and dynamic behavior of the reservoir models. Seismic reflection data, when included in the facies modeling, contributes additional information on the spatial distribution of the facies and leads to reservoir models which show better matches to production data. Including facies in a simultaneous inversion can have three additional advantages. Sharp boundaries in the modeled properties can be achieved. The models can incorporate a facies-dependent rock physics model linking the properties. Different levels of facies-dependent heterogeneity can be applied to the facies and the associated properties.
Geostatistical inversion algorithms represent each piece of input information, such as the seismic data, well logs and rock physics models, as probability density functions (pdfs). A Bayesian framework combines this information as “new evidence” along with the prior knowledge of the geology into a single set of “posterior pdfs”. A sampling scheme, such as a customized Markov Chain Monte Carlo sampling scheme, generates multiple plausible elastic, petrophysical and facies realizations. Analyzed together, this set of realizations can produce a quantitative measure of uncertainty. Statistical analysis and ranking of such realizations based on strategic criteria allows risk assessment in terms of mean, mode, and percentile (P10/P50/P90) models, which can then be used for further modeling through to such ends as flow simulation.
The presentation includes case study examples which illustrate the use of simultaneous inversion of facies and elastic properties within a geological context, made possible by using a geostatistical inversion algorithm in a Bayesian framework. The results show the range of situations where these more robust facies and property models can be used to benefit E&P decisions.