Gas Volume Fraction Measurement of Oil–Gas–Water Three-Phase Flows in Vertical Pipe by Combining Ultrasonic Sensor and Deep Attention Network

Abstract

Multiphase flow measurement is intimately linked with the production process optimization, production safety, and economic benefits. One of the challenging problems in flow parameter measurement is the gas volume fraction (GVF) measurements associated with the spatiotemporal structure of the flow patterns. This work aims to present an intelligent strategy to measure the GVF in the oil–gas–water three-phase flow with higher performance. The task is achieved by two contributions: first, a pulse transmission ultrasonic measurement system is designed that used a field-programmable gate array (FPGA) to control system; and second, a deep network architecture with attention mechanism combining convolutional neural network (CNN) and long short-term memory (LSTM) is proposed fed by data from the measurement system for real-time GVF prediction. The attention mechanism can help the network focus on the most informative regions of signal. The benefits of the proposed network are illustrated by comparison with the state-of-the-art theoretical model and other networks. The introduced strategy offers a new perspective on the flow parameter measurement of multiphase flow.