Skip to Content

Self-adjusting VFMs like our FLUX VFM are designed to provide the required accuracy without costly maintenance and interruptions over the lifetime of an asset.

Virtual Flow Metering (VFM) is an increasingly attractive method for estimating multiphase flow rates in oil & gas production systems. It is used to compute flow rates from readily available field measurements such as pressure and temperature. Since a VFM is purely software-based, it is easily integrated with other software systems, making the results from the VFM readily available for further production optimization analysis.

In order for a VFM to be a viable option, either as a standalone system or as a digital twin to a physical flow meter, the results from the VFM have to be sufficiently accurate. In most cases, a consistent full-scale error in the range of 5-10% for phase flow rates is required by the industry. The main challenge and key for a successful VFM solution is to keep the same level of accuracy for the lifetime of an asset, adjusting to changes as the field matures. Self-adjusting VFMs like our FLUX VFM are designed to provide the required accuracy without costly maintenance and interruptions over the lifetime of an asset.

One of the main challenges in consistently obtaining accurate phase flow rates using a first-principles transient flow simulator in a VFM, such as our transient FLUX Simulator, is to accurately determine how the inflow boundary condition changes over time, specifically for fluid composition and the fluid driving force (reservoir pressure in a well). A good representation of the inflow boundary enables the multiphase simulator to model the fluid flow through the well or pipeline, giving accurate phase flow rates at the desired measurement point.

Provided there are enough sensor values, sufficiently sensitive to changes in phase flow rates, boundary inflow parameters can be determined at any given point in time. In the FLUX VFM this is done using our FLUX Optimizer, which looks at how pressure and temperature changes across the well. We require a minimum of three pressure and temperature increases/decreases to resolve the three relevant inflow parameters: Gas-Oil-Ratio (GOR), Water-Cut (WC) and fluid driving force. In a well, this would typically be the pressure-drop across the choke and the pressure-drop and temperature-drop from bottom-hole sensor to the upstream choke sensor.

To provide even better accuracy in predicted inflow parameters, we have introduced the use of data analytics in addition to sensor values. We call it FLUX Analytics. Our data-driven model is trained to provide guidance to the FLUX VFM. This means that the result requirements from such a data-driven guidance-model is less severe than for a full data-driven VFM. We have done so because a first-principle VFM may find certain flow features useful, which may however not be interesting to the end-user. Worth to note is that this approach can also be used to create a VFM in case where the number of useful sensors is less than the required number of sensors. With that we are creating sufficient extra information to assure the required accuracy of our FLUX VFM.

FLUX VFM Architecture
FLUX VFM Architecture

Our FLUX VFM has successfully been used on a number of offshore and onshore wells. In the case of an onshore well project, the pressure-drop across the topside choke was insignificant due to the choke being fully open at all times. This meant that the pressure-drop was not sensitive to changes in flow. In addition, due to the small pressure-drop across the choke, the measurement error from the sensors were quite significant. Without any additional information this would have severely impacted the accuracy of the results.

It would have been required to have one of the inflow parameters fixed to get results of acceptable accuracy. However, the data models in our FLUX Analytics were able to capture various flow features such as the total volumetric flow rates based on sensor values. These features were used to guide the FLUX VFM in addition to the remaining sensor values. It meant that all inflow parameters could be derived and no parameters needed to be physically fixed. The overall full-scale error was reduced well within the 5-10% boundaries. An average full-scale error of less than 4% was even obtained for all phase flow-rates over a period of two months for one well, where the well experienced a 15-20% change in both GOR and WC.

Back to top