A virtual metering system is an alternative way to measure the flow rate of a well in real time. They can therefore improve the reliability and the effectiveness of the production back allocation process, providing redundancy to multi-phase flow meters (MPFMs) measures.
Scope of this work is to build and validate two different virtual metering systems highlighting their peculiarities and comparing their performance.
A virtual metering system can estimate the flow rates of each well of an asset by elaboration of pressure and temperature data coming from the field.
In both virtual metering systems, the core is a fluid-mechanical model of the production network of the same off-shore field, made of six wells flowing in two parallel lines. The model is run multiple times adjusting the wells flow rates according to two different minimization algorithms, to match the measured data as much as possible.
After validating the results, the performance of the tools has been compared against MPFMs, used as a common reference.
Both systems have been validated against officially allocated flow rates coming from MPFMs. At first, only pressure data have been used as inputs.
System number one, which exploits the Matlab minimization algorithm and an OLGA model of the network, showed great accuracy in the majority of the cases, with an error less than 5%, making it a great verification tool for measures coming from other instruments. Its simulation runtime, however, is still too long to make it usable in a real-time application scenario.
System number two relies instead on the gradient descent algorithm and on a GAP model of the network. This system is equipped with an automatic tool that can discard unreliable signals coming from damaged or out-of-calibration field sensors, while simulating. The resulting accuracy is acceptable, with an average error between 5% and 8% and the computational time is short enough to be used as a live measurement tool, in parallel to MPFMs.
The validation process has been repeated adding temperatures to pressures in the input dataset. No accuracy improvement has been shown by the new results for both systems, concluding that a leaner structure where only pressure is considered is to be preferred.
The two systems represent economic measurement tools for production allocation, showing a good ability in supporting field operation in case of instrument failure or unreliability.