A Machine Learning–Based Virtual Flow Meter for Continuous Estimation of Well Production Rates
Mohammad Amir Ashraff
Mohammad Amir Ashraff, Department of Data Science and Advanced Analytics, Hyderabad (Telangana), India.
Manuscript received on 09 January 2026 | First Revised Manuscript received on 19 January 2026 | Second Revised Manuscript received on 16 April 2026 | Manuscript Accepted on 15 May 2026 | Manuscript published on 30 May 2026 | PP: 1-9 | Volume-6 Issue-1 May 2026 | Retrieval Number: 100.1/ijpe.A192406010526 | DOI: 10.54105/ijpe.A1924.06010526
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Published by Lattice Science Publication (LSP). This is an open-access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Accurate estimation of liquid rate, gas rate, and water cut is essential for effective production surveillance, reservoir management, and operational decision-making in oil and gas assets. In most producing fields, direct production measurements are obtained through periodic well tests or selectively deployed multiphase flow meters, resulting in sparse temporal resolution, delayed detection of production changes, and limited field-wide visibility. Physics-based production models provide valuable engineering insight but require frequent calibration and often struggle to maintain accuracy under transient operating conditions and evolving reservoir behavior. These limitations motivate the use of data-driven approaches that leverage existing field instrumentation to deliver continuous production estimates. This paper presents a machine learning–based Virtual Flow Meter (VFM) for continuous estimation of oil, gas, and water production rates using routinely available operational measurements. High-frequency field data, including pressures, temperatures, choke settings, and, where applicable, lift-gas injection rates, are temporally aligned with historical well-test and laboratory measurements to construct reliable training datasets. Independent regression models are developed for oil, gas, and water rates, allowing each model to capture phasespecific sensitivities while maintaining physical consistency and avoiding reliance on explicit flow-regime identification or mechanistic multiphase correlations. The proposed VFM is deployed end-to-end on a commercial data analytics platform, enabling continuous ingestion of sensor data, real-time inference, performance monitoring, and periodic retraining. Model validation using historical field data demonstrates strong agreement between predicted and reference production values across all phases, indicating that the data- driven VFM provides reliable, meter-like production estimates. The results show that continuous production surveillance can be achieved without additional hardware or instrumentation, offering a scalable and cost-effective solution for large well portfolios. This approach supports proactive operational decision- making and enhances production visibility across modern oil and gas assets.
Keywords: Virtual Flow Meter, Machine Learning, Well Rate Estimation, Multiphase Flow, Production Surveillance.
Scope of the Article: Drilling Engineering