Multiphase flow metering – estimation of flow velocities of component phases using multimodal sensor suite

Abstract

There are many good alternatives to measure single-phase flow with high accuracy, but measuring multiphase flow is challenging. The sand detectors used on Equinor’s oil and gas installations use a piezoelectric sensor to measure the acoustic emission caused by sand production. This master thesis investigates if it is possible to make models for multiphase flow using a combination of acoustic emission sensors and other available sensors. If this is possible, then Equinor can use already installed sensors to estimate the multiphase flow on their oil and gas producing wells. This thesis shows that with a shallow neural network it is possible to create good models for gas and total flow rate with acoustic emission sensors and sensors to measure differential pressure over a Venturi. It was more challenging to make good models for the oil and water flow. To improve the results in this thesis it is recommended to try a deep neural network and to have higher focus on frequency analysis