Proper Selection of Multiphase Flow Correlations

Abstract

Abstract Multiphase pressure drop calculations represent a challenge for petroleum engineers. We normally rely on multiphase flow correlations to calculate pressure traverses through tubing from the sand face to the wellhead. Although there are many multiphase flow correlations covering most flow conditions; there are still no clear criteria to select appropriate correlations in absence of flowing gradient survey data. The objective of this work is to help petroleum engineers select the most accurate multiphase flow correlation(s) for any flow conditions, fluid properties, and well configuration. The method depends on analyzing a large database of pressure drop data from many wells. Our database is composed of more than 3,200 measured pressure points taken from survey and single point measurements on 879 wells representing large variations of flow conditions (oil rate ranges from very low rate to 31,000 BBL/D, water cut from 0 to 98%, gas-oil ratio from 0 to 20,000 SCF/STB, gas rate from very low rate to 200 MMSCFD, water gas ratio from 0 to 200 BBL/MMSCF, condensate gas ratio from 0 to 200 BBL/MMSCF for vertical, deviated and horizontal wells with different tubing sizes). We built models for all these wells and tested many available multi-phase flow correlations and recorded the error in their prediction against the actual measurements of pressure points. We then classified the large database into groups and found all multi-phase flow correlations within this group that have the lowest error. For every group, we can now identify the best correlation(s) with their estimated error value. We also defined a term to reflect the strength of the appropriate correlation(s) (how many times this particular correlation was best in each group of data) to help engineers select appropriate correlation for each well and its flowing conditions. The strength term for all groups varies from 100 % to 30% which indicates that for several groups we can now select the multiphase flow correlation with significant accuracy. The expected error for all correlations is also reported. The parameters that mostly affect each correlation are also highlighted so engineers would know which parameters need to be accurate in using the recommended correlations.