A Data Science Application for Validating Multiphase Measurement by Separator Data

Abstract

Multiphase flow refers to the simultaneous flow of two or more immiscible fluids with distinct physical properties through a pipeline. In the oil and gas industry, these flows typically comprise oil, gas, and water, which are separated on production platforms by separator vessels to allow accurate individual flow rate measurements. However, due to the size and weight of such vessels, there is growing interest in using multiphase flow meters (MPFM) as alternatives for field testing. This paper applies a calibration factor methodology to validate MPFM measurements using reference data from single-phase flow meters positioned downstream of the test separator. The calibration factor is calculated as the ratio between the cumulative volume measured by the single-phase meters and that measured by the MPFM. Field data from 27 tests conducted in two campaigns across seven wells on a Brazilian offshore platform were analyzed using a Python-based adaptation on Gustavsen's algorithm. The results showed that while the MPFM tends to estimate the total liquid flow accurately, significant deviations occur when measuring individual phases, especially oil. Calibration factors were analyzed over time and across wells, with convergence and percentage error criteria used to assess measurement reliability. A Kruskal-Wallis statistical test revealed that well-specific properties influence MPFM accuracy, suggesting the need for calibration protocols by well rather than by device. In a second campaign, adjustments in sensor acquisition frequency significantly improved calibration factor accuracy, demonstrating that MPFM performance can be enhanced through technical refinements. Additionally, a complementary 50/50 calibration and validation approach was implemented, in which half of each dataset was used to estimate the calibration factor and the remaining half to evaluate its performance. The Mean Absolute Percentage Deviation (MAPD) confirmed that calibration effectively reduced deviations, with oil MAPD decreasing from 58.36% to 8.74%, water from 10.46% to 1.69%, and gas from 29.36% to 24.92%.