Modeling of two-phase gas-oil flow in a vertical pipe with viscous oil

Abstract

Most of the physical models that are used for modeling two phase gas oil flows are derived from experiments with less viscous oil. This paper has studied the applicability of these models in commercial codes for flows with high viscous oil. Statoil ASA has provided experimental data of two phase gas oil flows with high viscous oil in vertical pipes with large pipe diameter. The dynamic multiphase flow simulator OLGA has been used for predicting pressure drops, liquid holdups and flow patterns. In addition, a point model provided by Norwegian University of Science and Technology has been utilized in order to see if pressure drop and liquid holdup could be captured by a simple, steady state model. Deviations between predictions and experimental data have been studied, with particular attention on oil viscosity.Poor predictions were observed. For oil viscosity exceeding three hundred centipoise, OLGA consistently overpredicted pressure drops. A new method for determining flow regimes was established from normal probability density plots of line fraction measurements. Slug flow was substituted by intermittent flow, as previous experiments that were conducted on the same facility indicated that the characteristic shape of Taylor bubbles does not exist in large diameter pipes. Transitions in the experimental flow pattern map showed characteristic shifts compared to results from OLGA and literature. The worst predictions occurred for liquid holdups, which is due to large uncertainties in the conversion of one dimensional line fraction measurements into real holdup values. For most cases, the point model gave more accurate predictions of total pressure drops than OLGA. This phenomenon was ascribed to the flow conditions of which the empirical friction models in the point model have been derived. Other recent studies found the pressure drop model of Hagedorn and Brown to be most accurate, but it was one of the worst performing models for current experimental data. This disagreement is described with the scale up effects of large pipe diameters. Proven existence of scale up effects created difficulties in separating the direct influence of high liquid viscosity. Further studies are necessary in order to establish flow models that are able to capture the effects of high viscous oils.