Uncertainty Analysis of Oil Sands Reservoirs Using Features in Metric Space

Abstract

In oil sands reservoirs, it is difficult to predict the production reliably because a steam chamber continues to expand with time and the affected area keeps changing. This article proposes a new approach for history matching and uncertainty analysis in oil sands reservoirs. The first step is to set an area affected by the expansion of the steam chamber. Then some features are extracted and plotted in metric space using 2D discrete Fourier transform and principal component analysis. Several realizations are selected based on the same interval sampling and ranked with similar levels in production curve. The sampling weights are assigned proportional to the rankings and one realization is selected accordingly. A flow simulation performs the nearest point (realization) around the selected one in metric space. The most dissimilar realization is replaced in the candidates. The previous step is repeated until the objective function meets a stop criterion. In the result of uncertainty analysis, it could lead to a fast convergence and preserve geological reality. The mean facies map of the candidates generates successfully the position of the shale barrier in the reference field.