Reservoir Modeling and Volumetric Uncertainty for Field Development Planning: An example of Upper Permian Formation in Su6 Area, Ordos Basin, West-Central China

The geological model is a set of multiple geostatistical realizations that incorporate plenty of uncertainties in the procedure of reservoir modeling. Integrated reservoir modeling procedure and ultimate volume calculation has... [ view full abstract ]

The geological model is a set of multiple geostatistical realizations that incorporate plenty of uncertainties in the procedure of reservoir modeling. Integrated reservoir modeling procedure and ultimate volume calculation has introduced significant uncertainties to guiding decisions for field development planning. A multi-step modeling procedure that incorporates structural modeling, lithofacies modeling and petrophysical modeling is applied to the Su6 area discovery, a strong heterogeneity, and tight gas field in the central Ordos basin: (1)Structural modeling accounts for the gross rock volume (GRV) uncertainty by changing reservoir thickness. In this process, stochastic error surfaces are generated for each realization, which indicates a high confidence level in the structure of 3D grid. (2)Lithofacies modeling determines the reservoir volume, i.e. ‘gas container’. Their spatial distribution and proportions are uncertain to some degree, which will highly influence the subsequent population of reservoir properties within the framework of lithofacies models. (3)Petrophysical modeling was proceeded in two steps: porosity and permeability modeling, property cut-offs and net to gross (N/G) ratio modeling. Also net reservoir volume uncertainty derived from petrophysical modeling varies a lot with each realization. (4)Gas saturation (Sg) modeling and gas initially in place (GIIP) calculation. Due to the limitations and inapplicability of Archie’s formulas, calibration from core to well-logs, Sg model turns out to be especially uncertain or inaccurate. Besides, different stochastic algorithms applied to modeling such as Sequential Gaussian simulation or multiple-point geostatistics simulation, will also generate different realizations. Based on the multiple realizations of each procedure and Monte-Carlo sampling method, we will obtain the probability distribution curves of GRV, effective porosity, N/G, Sg and GIIP. As a result, the distribution curve of GIIP mapping the P10, P50, P90 values shows the reserve range from 0.8×1012 m³ to 1.13×1012m³. Sensitivity analysis displayed in a tornado chart reveals the key contributors to final GIIP calculation are N/G and effective porosity, then the Sg.