Multi-task learning for virtual flow metering

作者：

Highlights：

• Large-scale data-driven virtual flow meter study with 55 wells from four assets.

• Multi-task learning architectures better adhere to physical expectations.

• Sharing data can improve performance by 25%–50% in challenging cases.

摘要

•Large-scale data-driven virtual flow meter study with 55 wells from four assets.•Multi-task learning architectures better adhere to physical expectations.•Sharing data can improve performance by 25%–50% in challenging cases.

论文关键词：Neural network,Shared parameters,Multi-task learning,Virtual flow metering,Multiphase flow

论文评审过程：Received 17 March 2021, Revised 31 August 2021, Accepted 31 August 2021, Available online 2 September 2021, Version of Record 20 September 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107458