Integrated Semi-Supervised Clustering Method and Its Application in Rock-Typing in AA Reservoir

Abstract

Rock-typing is complicated and critical for numerical simulation. Therefore, some researchers proposed several clustering methods to make classification automatic and convenient. However, traditional methods only focus in specific area such as lithofacies or petrophysical data instead of integrated clustering. Besides, all the clustering method are related to classification interval determined subjectively. Therefore, a new clustering method for rock-typing integrated different disciplines is critical for modelling and reservoir simulation. In this paper, we proposed a novel semi-supervised clustering method integrated with data from different disciplines, which can divide rock type automatically and precisely. Considering AA reservoir is a porous carbonate reservoir with seldom fracture and vug, FZI (Flow Zone Indicator) and RQI (Reservoir Quality Index) is utilized as the corner stone of the clustering method after collection and plotting for porosity and permeability data for cores from AA reservoir. Then lithofacies, sedimentary facies and petrophysical data are applied as constraints to improve FZI method. Hamming distance and earth mover distance are imported to build integrated function for clustering method. Finally, based on output results of integrated clustering method from experimental data, grid properties of model in Petrel software are imported as the input parameter for further procession. Therefore, saturation region for numerical simulation built by rock-typing is constructed. The results show that new method could make classification accurately and easily. History matching results for watercut indicate that new saturation regions improve the numerical simulation performance.