Improved Procedures for Estimating Uncertainty in Hydrocarbon Recovery Predictions

by
Jirawat Chewaroungroaj, Ph.D.

University of Texas at Austin, 2000
Supervisor: Larry W. Lake

Uncertainty stems from our failure or inability to measure, represent, or understand all of the features of a reservoir. Development of different scenarios in predictions of hydrocarbon recovery is the manifestation of the lack of information on or uncertainty about reservoir features. A statistical treatment that recognizes both the lack of knowledge and the uncertainty of the parameters involved in the forecast of reservoir performance is desirable. Monte Carlo simulation has become a typical method for assessing the uncertainty in the hydrocarbon recovery predictions. This method often requires numerous flow model calculations and the associated computational burden can be prohibitive. The thrust of this study is to develop procedures to perform uncertainty estimation with less computational effort but with, it is presumed, only a small loss of accuracy.

This study demonstrates alternatives to Monte Carlo simulation that quantitatively estimate uncertainty in a specific hydrocarbon recovery prediction. A simple primary recovery of a slightly compressible oil above water table is chosen as the study process. An approximate analytical technique based on Taylor's series expansion, the first-order approximation, is presented. This simple approach considers the effects of both sensitivity and uncertainty on variability of input variables. Two experimental design techniques, the Box-Behnken and Taguchi approach, are employed and results are compared to Monte Carlo simulation. The Box-Behnken experimental design can provide a reasonably accurate uncertainty estimation of hydrocarbon recovery with fewer simulation runs than the Monte Carlo simulation. The Taguchi approach underestimates the uncertainty in hydrocarbon recovery in this case study. Application of response surface method using the experimental designs to predict oil recovery and associated uncertainty is illustrated.

The use of multiple processors reduces the turnaround time of the Monte Carlo simulation runs. The partial simulation technique can be used to determine the workload distribution to a cluster of computers for efficiency improvement. The combination of these improved procedures and the multiple computing units will reduce the computational effort in estimating uncertainty of hydrocarbon predictions.

Back to theses index