Clustering of Geological Models for Reservoir Simulation Studies in a Visual Analytics Framework
Z. Sahaf, H. Hamdi, F. Maurer, L. Nghiem and M. Costa Sousa
Event name: 78th EAGE Conference and Exhibition 2016
Session: Quantifying and Managing Uncertainty in Reservoir Modelling
Publication date: 31 May 2016
Info: Extended abstract, PDF ( 1.2Mb )
Price: € 20
In this paper, we introduced a new visual analytics framework to select a few representative models from an ensemble of geostatistical models that can represent the overall production uncertainty. To achieve this purpose, a new block based similarity metric is defined based on mutual information. This metric is computed based on static geological properties and helps to identify geological models with similar flow simulation results. In the next step, utilizing the computed similarity values, a customized multi attribute clustering algorithm is applied on a set of geological models. One model is selected from each cluster randomly that results in having a few representative geological models. The whole process is implemented in a visual analytics framework. The proposed workflow was exemplified using some datasets generated from various geostatistical facies realizations using different variogram correlation lengths. The results on the case studies show that our technique is 70% accurate and much more time efficient in comparison to the existent techniques. The method is being enhanced for more accurate clustering.