Quantification of uncertainty during history matching

This study proposes a new, easily applied method to quantify uncertainty in production forecasts based on reservoir simulation. The new method uses only observed data and mismatches between simulated values and observed values as history matches of observations progress to a final "best" match. The method is applicable even when only limited information is available from a field. Previous methods suggested in the literature require more information than our new method. Quantifying uncertainty in production forecasts (i.e., reserve estimates) is becoming increasingly important in the petroleum industry. Many current investment opportunities in reservoir development require large investments, many in harsh exploration environments, with intensive technology requirements and possibly marginal investment indicators. Our method of quantifying uncertainty uses a set of history-match runs and includes a method to determine the probability density function (pdf) of future oil production (reserves) while the history match is evolving. We applied our method to the lower-Pleistocene 8-Sand reservoir in the Green Canyon 18 field, Gulf of Mexico. This field was a challenge to model because of its complicated geometry and stratigraphy. iv We objectively computed the mismatch between observed and simulated data using an objective function and developed quantitative matching criteria that we used during history matching. We developed a method based on errors in the mismatches to assign likelihood to each run, and from these results, we determined the pdf of reservoir reserves and thus quantified the uncertainty in the forecast. In our approach, we assigned no preconceived likelihoods to the distribution of variables. Only the production data and history matching errors were used to assess uncertainty. Thus, our simple method enabled us to estimate uncertainty during the history-matching process using only dynamic behavior of a reservoir.