The evolution of well testing Taking production from the typical multiphase combination of oil, gas, water, and solids such as sand or asphaltenes to exacting measurement practices has always been challenging for operators. When those elements reach the inlet separator, and the mixture splits into single phases, this is where accurate measurement of flow rates becomes, and will always be, critical to the economic and operational success of the field. Traditionally, well testing was the primary method for estimating flow rates. This involves directing the well stream into a test separator, splitting it into oil, gas, and water. These streams are then measured by single phase meters at the separator outlet, each requiring a separate flowline for individual routing and testing while avoiding an entire field shutdown. In the 90’s, multiphase flow meters (MPFM) emerged as an alternative to well testing. But despite the capacity for real-time flow rate information, there were significant capital costs – mainly down to the need for separate flowlines and separators and expensive maintenance interventions. A new way for a new era Today, we’re in a world of rich data and abundant information. It’s now possible to leverage the power of insight, vastly improving performance and profit. Virtual Flow Metering (VFM) estimates oil, gas, and water flow rates using field data such as pressure, temperature, and choke position combined with hydrodynamic multiphase models and reconciliation algorithms. There is no requirement for additional hardware either, meaning a reduction in capital expenditure. The adaptability of VFMs means they can provide real-time flow rate estimates and adjust them according to changing conditions, as opposed to assuming a constant flow rate, as was the case with the MPFM. What’s more, VFM is not a standalone technology – it can utilize the data provided by MPFM systems through integration. Combing physics and machine-learning-based approaches VFM approaches can be physics-based or machine-learning-based, or a combination of the two. Physics based VFM relies on multiphase flow simulations grounded in the well-established principles of thermodynamics, fluid dynamics, and optimization techniques. It requires extensive domain knowledge and can be computationally intensive. On the other hand, machine-learning-based VFM uses algorithms to find relationships between sensor data and target variables in historical datasets. It requires less deep physical knowledge but does depend on high-quality data and may struggle with changes in operational conditions. A hybrid VFM solution combines physics-based models with machine learning, leveraging strengths in both. Machine learning is applied to synthetic data generated from validated physics-based models, addressing gaps in training data and simulating various conditions and scenarios. We’ve found that a hybrid method offers clear benefits over purely model-based or data-driven approaches, enhancing accuracy and adaptability in flow rate estimation. Taking Intelligent Action in flow metering With our history in the industry, we’re in the unique position of understanding decades of innovation in flow metering. So much so we’ve developed VFM that combines physics with machine learning – unlocking the power of data, digital technology, and automation control at a fraction of the cost. It creates excellence in the industry that produces better profitability, performance, and reliability in oil and gas production, freeing up decision-makers to take Sustainable Intelligent Action across their operation. Learn more about our intelligent measurement and flow metering solutions.
