Modelling gas condensate viscosity below the dew point using Fuzzy Logic approach

bank of condensate can rapidly build up around a producing well, when bottom hole flowing pressure falls below the dew point in a depleted gas condensate reservoir. This condensate bank will directly affect well performance as it impedes gas flow to the surface. Forecasting production as well as optimising future recoveries in such reservoirs are highly desirable. For the purpose of accurate forecasting it is unavoidable to determine accurate viscosity of condensate phase below the dew point.
Measurement of condensate viscosity needs highly sophisticated equipment, which is expensive and require high level of skills. Hence, numerical approaches have become a commonplace in the oil and gas industry. The objective of this study is to develop a numerical model using Takagi-Sugeno-Kang (TSK) fuzzy inference system to predict condensate viscosity in depleted gas condensate reservoirs below the dew point.
TSK fuzzy model has been developed and tested using 326 series of condensate viscosity experimental data sets collected from existing literature. Acquired input data sets organized in matrix form, which enables optimum number of clusters to be determined using Calinski-Harabasz cluster evaluation method. Knowing the optimum number of clusters (rules), Gaussian membership functions were used to relate the degree of membership of the input/output data sets in different clusters.
Subsequently, the condensate viscosity functions are generated in terms of pressure, temperature and solution gas to oil ratio using least-square approximation method.
Comparative statistical and graphical error analysis have been carried out between new TSK fuzzy model and the most frequently used gas-saturated-oil viscosity correlations. The outcome of this study show that the developed fuzzy model outperforms the existing correlations with lowest Root Mean Square Error (RMSE) of 0.0649 and Mean Absolute Error (MAE) of 0.055.