Uncertainty Quantification in Reservoir Performance in Metric Space - Application to a West-Africa Deepwater Turbidite Reservoir.

@inproceedings{ScheidtGM2009,
 abstract = { Stochastic simulation allows rapid generation of multiple realizations of spatial variables. Quantifying uncertainty on responses resulting from those multiple realizations would require the evaluation of a transfer function on every realization. This is not possible in real applications, where one transfer function evaluation may be very time consuming. One must therefore select a few representative realizations for transfer function evaluation and then derive the production statistics of interest (typically the P10, P50 and P90 quantiles of the response). In this paper, we propose to select realizations using a distance-based kernel clustering technique (DKM), and compare it to the traditional ranking technique. However, by selecting only a few realizations one may risk generating unreliable P10, P50 and P90 estimates compared to a larger set of realizations. Therefore, we propose a method to quantify confidence intervals for the estimated quantiles. Our approach is to use the parametric bootstrap technique, which allows the evaluation of the variability of the statistics obtained from uncertainty quantification and the construction of confidence intervals. We compare confidence intervals when using traditional ranking technique and DKM. A case study is presented on a deepwater turbidite offshore reservoir in west Africa. The reservoir is modeled using 4 facies whose spatial distribution is uncertain due to uncertain facies proportions, location, etc. Alternative training images are defined to capture the spatial uncertainty of the prior model. Then, many realizations are generated with these training images as input. We show that quantification of uncertainty on a well selected subset of realizations results in similar statistics as the uncertainty of the full set. In addition, the results show that for the same number of transfer function evaluations, the DKM method is more accurate in reproducing the statistics of the full set, and has equal or smaller confidence interval compared to ranking. The applied workflow thus produces a prediction of P10, P50 and P90 quantiles, and associated confidence intervals. },
 author = { Scheidt, Celine AND Caers, Jef },
 booktitle = { Proc. 29th Gocad Meeting, Nancy },
 title = { Uncertainty Quantification in Reservoir Performance in Metric Space - Application to a West-Africa Deepwater Turbidite Reservoir. },
 year = { 2009 }
}