Data Driven Versus Transient Multiphase Flow Simulator for Virtual Flow Meter Application

This study aims to evaluate two independent approaches of Virtual Flow Meter (VFM) i.e., using Transient Multiphase Flow Simulator (TMFS) and data-driven using Diverse Ensemble Learning Neural Network (DELNN). The main objective of using the Virtual Flow Meter (VFM) developed from this study is to implement in real time as a mean of troubleshooting and validating the measurement provided by a physical Multiphase Flow Meter (MPFM) for well testing operation. The result of the study showed both VFM flow rate estimates were less than 10% of full-scale error for both oil and gas flow rates compared to the measured flow rate respectively. Additionally, both VFM also independently managed to track a similar trend of deviation in gas flow rate which help to identify failure in the Multiphase Flow Meter (MPFM) internal measurement devices. The result of the study proved that by employing two independent VFM approaches in parallel, we could position VFM with higher confidence as a reliable solution either as a backup or as a mean of troubleshooting solution to physical MPFM as well as an analytic tool to plan well testing procedure.