An Estimation of Water-Oil Relative Permeability Full Curves by ANFIS

Abstract

Adaptive neuro-fuzzy inference system is an intelligent nonlinear, multivariable technique. In this study, an adaptive neuro-fuzzy inference system was used to identify complex relations between water-oil relative permeability curves and rock and fluid properties. Sixty-seven relative permeability curves from Iranian carbonate and sandstone reservoirs were used in this study. An adaptive neuro-fuzzy inference system was then used to predict un-normalized relative permeability full curves as a function of different rock and fluid properties. High correlation coefficients (R 2) of 0.88 and 0.94 were obtained for oil and water relative permeability curves, respectively. Water relative permeability curves also show no dependency to wettability index of the core and fluid system while oil curves show weak dependency. Adaptive neuro-fuzzy inference system models showed very promising to obtain real water-oil relative permeability full curves when the required core and fluid properties are available. Finally, the effect of porosity, initial water saturation, and wettability on relative permeability curves was investigated by constructed models. Relative permeability of both phases increased with porosity and showed decreasing with initial water saturation. Wettability has no considerable effect on relative permeability curves.