Virtual flow meter based on artificial intelligence is implemented in oil wells

According to Energy Press, citing the Oil Engineering and Development Company, Saeed Saviz, referring to the importance of the virtual flow meter project based on artificial intelligence, whose contract was signed yesterday (1st of December), said: Accurate measurement of production flow in oil wells has always been one of the technical and economic challenges of the oil industry. The purpose of implementing the virtual flowmeter design and construction contract is to move towards software-oriented, safe and data-oriented solutions that can provide users with reliable information at a lower cost and higher flexibility.
He added: In the past years, virtual flow meter technology (MPFM) was known as the reference solution for measuring multiphase flow, but the reliance of this technology on nuclear resources, safety considerations, strict operational requirements and high costs have caused the world’s oil industry to gradually move away from it.
Localization path and future horizon
The head of research and development affairs of Oil Engineering and Development Company said: the virtual flow meter based on artificial intelligence can estimate the flow rate and flow behavior with acceptable accuracy only by using the operational data of the well. This approach indicates a paradigm shift from hardware-oriented to algorithm-based intelligence and data analysis.
Saviz stated that according to the terms of this contract, the designed virtual flowmeter will be able to be used in wells equipped with pumps and wells without pumps, and clarified: the artificial intelligence algorithms of this system are trained according to the type of well. Differences in flow behavior, production conditions, and input data are included in the models to maintain the accuracy of calculations in different scenarios.
According to him, data such as pressure, temperature, pump power consumption and other available parameters form the main input of artificial intelligence models and the system improves its performance over time and by learning from real data.