Next Generation Geological Modeling for Hydrocarbon Reservoir Characterization - Advances in Data, Methods, Models and Their Applications in Geoscience - page 236

user's discretion only before incorporation into the reservoir simulator. Models, generated
in such fashion, retain the maximum available resolution and information density, limited
only by the resolution of underlying data and structural continuity.
Regardless of the modeling approach, geoscientists and engineers often select diverse
geomodel realizations such that the reservoir simulation outcome will cover a sufficiently
large range of uncertainty to approximate the reservoir recovery forecast statistics
throughout the asset lifecycle. One of the differentiating attributes of next-generation
reservoir characterization is to integrate the reconciliation of geomodels with well-
production and seismic data, with dynamic ranking and selection of representative model
realizations for reservoir production forecasting. We outline and validate the evolving
technology for Quantitative Uncertainty Management that seamlessly interfaces the
DecisionSpace  Desktop, VIP ® and/or Nexus ® reservoir simulators. The highest available
adherence to geological detail with respect to structural features that control depositional
continuity ( e.g. facies) is maintained through implementation of advanced model
parameterization, based on Discrete Cosine Transform (Rao and Yip, 1990), an industrial
standard in image compression. The AHM algorithm utilizes a highly efficient derivative of
sequential (Markov chain) Monte Carlo sampling, where the acceptance rate is increased
and computational effort reduced, by the utilization of streamline-based sensitivities.
By its nature, the probabilistic HM workflows render multiple equally probable but non-
unique realizations of geological models that honor both, prior constraints and production
data, with associated uncertainty. Throughout the inversion, however, some model
realizations may have created non-geologically realistic features and these models are
inadequate for forecasting of recovery performance. We outline the workflow for dynamic
quantification production uncertainty that utilizes rapid streamline simulations to calculate
data-pattern dissimilarities, Multi Dimensional Scaling to correlate dynamic model
responses with pattern dissimilarities and kernel-based clustering methods to intelligently
identification and ranking of geo-models, representative in forecasting decisions. When
integrated into the DecisionSpace  Desktop suite of reservoir characterization tools, such
technology will assist in Smart Reservoir Management and decision making by combining
multiple types and scales of data, honoring most first order effects, capturing a full range of
outcomes and reducing analysis and decision time.
4. Acknowledgement
The authors would like to acknowledge Halliburton/Landmark for the permission to
publish this text.
5. References
Alvarado, V. & Manrique, E. (2010). Enhanced Oil Recovery: Field Planning and
Development Strategies, Elsevier, MO, 208 p.
Borg, I. & Groenen, P.J.F. (2005). Modern Multi-dimensional Scaling: Theory and Applications ,
2 nd Ed., Springer, NY, pp. (207-212).
Caers, J. (2005). Petroleum Geostatistics , SPE Primer Series, Richardson, TX, 88 p.