The effects of incorporating dynamic data on estimates of uncertainty

Abstract

Petroleum exploration and development are capital intensive and smart economic decisions that need to be made to profitably extract oil and gas from the reservoirs. Accurate quantification of uncertainty in production forecasts will help in assessing risk and making good economic decisions. This study investigates the effect of combining dynamic data with the uncertainty in static data to see the effect on estimates of uncertainty in production forecasting. Fifty permeability realizations were generated for a reservoir in west Texas from available petrophysical data. We quantified the uncertainty in the production forecasts using a likelihood weighting method and an automatic history matching technique combined with linear uncertainty analysis. The results were compared with the uncertainty predicted using only static data. We also investigated approaches for best selecting a smaller number of models from a larger set of realizations to be history matched for quantification of uncertainty. We found that incorporating dynamic data in a reservoir model will result in lower estimates of uncertainty than considering only static data. However, incorporation of dynamic data does not guarantee that the forecasted ranges will encompass the true value. Reliability of the forecasted ranges depends on the method employed. When sampling multiple realizations of static data for history matching to quantify uncertainty, a sampling over the entire range of realization likelihoods shows larger confidence intervals and is more likely to encompass the true value for predicted fluid recoveries, as compared to selecting the best models.