Application of GK Fuzzy Clustering Method for PVT Analysis of Gas Condensate Reservoirs

In this work Gustafson-Kessel (GK) FCM algorithm was implemented and its effectiveness in handling high dimensional data was revealed. This algorithm associates each data point in the data set with every cluster using an optimized membership function. GK forms a generalization of the FCM algorithm by utilizing the Mahalanobis distance for non-spherical clusters. In this clustering algorithm, components are placed in the hyper-component along with simultaneous calculation of critical and thermo-physical properties. Four case studies were selected for the characterization in PVT analysis of gas condensate reservoir fluids. The mixture composition and properties of the gas condensate samples of reliable published data are used. The automatic placement of components in each group is consistent with previous schemes those have highly heuristic nature of pseudo-component generation. The perfect agreement between detailed and clustered PVT analysis, shows good predicting capability of this clustering algorithm in mixture characterization and pseudo-component generation to simulate thermodynamic equilibrium and volumetric behavior in PVT experiment designed for gas condensate reservoir including prediction of condensed liquid dropout, densities, viscosities and saturation pressure.