Integration of geostatistical methods and clustering analysis for the selection of representative geological cases

Summary The characterization and modelling of the hydrocarbon reservoirs include the integration of data from different sources and scales, which leads to considering an important number of geological uncertainties. Additionally, during reservoir modelling, it is possible to consider several scenarios and strategies that increase the number of geological variables, workflows, and uncertainty analysis thereof. This paper proposes a simple way to select representative geological workflows based on hydrocarbon volume percentile clustering, thereby reducing the number of cases to be simulated during history matching. This methodology is applied in a developed conventional field that has 40 wells and started oil production since the 2000s. The reservoir 3D grid of this field includes structural, facies, rock types and petrophysical property modelling strategies, which were evaluated by setting up different modelling algorithms, geological variables, and multi-variable dependencies. Monte Carlo analysis of the different modelling strategies gave volumetric cases, which are inputs for the clustering analysis. The centroids of each cluster were selected as the most representative workflow to be proposed to the dynamic simulation. This permitted to perform geological and dynamic reservoir uncertainty modelling in an efficient manner.