2021 Virtual Undergraduate Research Symposium

2021 Virtual Undergraduate Research Symposium

Machine learning based two-phase flow monitoring using DAS

Machine learning based two-phase flow monitoring using DAS

PROJECT NUMBER: 58 | AUTHOR: Ana Garcia-Ceballos​, Geophysics

MENTOR: Ge Jin, Geophysics

ABSTRACT

The success of a producing reservoir is achieved by understanding reservoir performance through production flow monitoring along a well. Recently, more hydrocarbon industries have become interested in utilizing distributed acoustic sensing (DAS) for borehole sensing applications. This technology uses optical fiber as a sensor, providing dense data spacing, making it a reliable and cost-effective method. Due to the large amounts of data recorded, DAS data processing is challenging. This study investigates machine learning application to automate data processing of DAS data for multiphase flow analysis. The Reservoir Characterization Project consortium at Colorado School of Mines previously built a vertical flow loop with optical fiber wrapped along a 7-meter PVC pipe. Water and air sources are used to create a two-phase flow. Analyzed DAS data was collected using an OptaSense ODH 3.1 interrogator unit for the duration of 1 minute. The data was processed using frequency-time analysis following data preconditioning. Next, unsupervised k-means clustering machine learning algorithm was applied to analyze the fluid flow. Two out of four groups of clusters characterized by the method move with the flow air phase. The proposed data processing algorithm automates multiphase flow characterization to estimate slugging velocity and can be scaled for wellbore usage as well as for other applications such as pipeline monitoring.

PRESENTATION

AUTHOR BIOGRAPHY

Ana Garcia-Ceballos is a senior undergraduate in the Geophysical Engineer program in the Geophysics Department at Colorado School of Mines. She utilizes machine learning and fiber optics technology to monitor and characterize the flow of fluid mixtures within a pipe. She is interested in how bubbles behave within a fluid and how they can potentially damage infrastructure and production while in transport. Her interest in utilizing fiber optics also lies in the monitoring of natural hazards with submarine cables, as well as the production of gas hydrates.

1 Comment

  1. This is a really neat project. I like the use of machine learning for a real-world application rather than just a computer application.

Share This