Quantification of reservoir uncertainty for optimal decision making

A reliable estimate of the amount of oil or gas in a reservoir is required for development decisions. Uncertainty in reserve estimates affects resource/reserve classification, investment decisions, and development decisions. There is a need to make the best decisions with an appropriate level of technical analysis considering all available data. Current methods of estimating resource uncertainty use spreadsheets or Monte Carlo simulation software with specified probability distributions for each variable. 3-D models may be constructed, but they rarely consider uncertainty in all variables. This research develops an appropriate 2-D model of heterogeneity and uncertainty by integrating 2-D model methodology to account for parameter uncertainty in the mean, which is of primary importance in the input histograms. This research improves reserve evaluation in the presence of geologic uncertainty. Guidelines are developed to: (a) select the best modeling scale for making decisions by comparing 2-D vs. 0-D and 3-D models, (b) understand parameters that play a key role in reserve estimates, (c) investigate how to reduce uncertainties, and (d) show the importance of accounting for parameter uncertainty in reserves assessment to get fair global uncertainty by comparing results of Hydrocarbon Initially-in-Place (HIIP) with/without parameter uncertainty. The parameters addressed in this research are those required in the assessment of uncertainty including statistical and geological parameters. This research shows that fixed parameters seriously underestimate the actual uncertainty in resources. A complete setup of methodology for the assessment of uncertainty in the structural surfaces of a reservoir, fluid contacts levels, and petrophysical properties is developed with accounting for parameter uncertainty in order to get fair global uncertainty. Parameter uncertainty can be quantified by several approaches such as the conventional bootstrap (BS), spatial bootstrap (SBS), and conditional-finite-domain (CFD). Real data from a large North Sea reservoir dataset is used to compare those approaches. The CFD approach produced more realistic uncertainty in distributions of the HIIP than those obtained from the BS or SBS approaches. 0-D modeling was used for estimating uncertainty in HIIP with different source of thickness. 2-D is based on geological mapping and can be presented in 2-D maps and checked locally.