Citation: Arief, H.A.; Thomas, P.J.;
Constable, K.; Katsaggelos, A.K.
Towards Building a Distributed
Virtual Flow Meter via Compressed
Continual Learning. Sensors 2022, 22,
9878. https://doi.org/10.3390/
s22249878
Academic Editors: M. Jamal Deen,
Subhas Mukhopadhyay, Yangquan
Chen, Simone Morais, Nunzio
Cennamo and Junseop Lee
Received: 11 November 2022
Accepted: 12 December 2022
Published: 15 December 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
Towards Building a Distributed Virtual Flow Meter
via Compressed Continual Learning
Hasan Asy’ari Arief
1,2,
* , Peter James Thomas
1
, Kevin Constable
3
and Aggelos K. Katsaggelos
2
1
NORCE Norwegian Research Centre AS, 5008 Bergen, Norway
2
Department of Electrical and Computer Engineering, Northwestern University, Evanston, IL 60208, USA
3
Equinor AS, 5254 Bergen, Norway
* Correspondence: hasv@norceresearch.no
Abstract:
A robust–accurate estimation of fluid flow is the main building block of a distributed
virtual flow meter. Unfortunately, a big leap in algorithm development would be required for this
objective to come to fruition, mainly due to the inability of current machine learning algorithms
to make predictions outside the training data distribution. To improve predictions outside the
training distribution, we explore the continual learning (CL) paradigm for accurately estimating
the characteristics of fluid flow in pipelines. A significant challenge facing CL is the concept of
catastrophic forgetting. In this paper, we provide a novel approach for how to address the forgetting
problem via compressing the distributed sensor data to increase the capacity of the CL memory bank
using a compressive learning algorithm. Through extensive experiments, we show that our approach
provides around 8% accuracy improvement compared to other CL algorithms when applied to a
real-world distributed sensor dataset collected from an oilfield. Noticeable accuracy improvement
is also achieved when using our proposed approach with the CL benchmark datasets, achieving
state-of-the-art accuracies for the CIFAR-10 dataset on blurry10 and blurry30 settings of 80.83% and
88.91%, respectively.
Keywords: virtual flow meter; continual learning; distributed acoustic sensing
1. Introduction
The distributed virtual flow meter can provide a game-changing functionality in the
oil and gas industry, and other industries requiring accurate characterization of fluid flows.
With a distributed measurement capability applied to oil and gas production, it is possible
to detect and locate water/gas breakthrough, monitor fractures and pressure drops as they
occur, and perform a noninvasive well integrity inspection [1].
Physical devices, i.e., optical flow meters, optical tomography, gamma densitometry,
and test separators, can provide high-accuracy fluid flow characterization. Unfortunately,
such devices are either radioactive, require an invasive operation, or do not provide
distributed measurement. Figure 1 depicts an illustration of several different flow meter
devices within an oil rig. The test separator works by physically separating fluid contents
inside an oil pipe, and then measures the volume of water, oil, and gas individually. This
type of measurement is the most accurate compared to other existing devices. However, the
test separator operation requires a redirection of the fluid flow from normal operation to the
separator box; thus, it is only used for getting a snapshot in time of the fluid composition.
Tomography devices on the other hand, are expensive and require active reading by
beaming ultrafast X-rays pulse to measure flow changes and velocity measurement [
2
].
Another family of flow meter devices are optical flow meters; they measure the fluid
flow using the speed of sound obtained from the acoustic waves of turbulent flows. The
optical flow meters have been adopted as an industry standard for cost-effective flow meter
devices. However, they only provide point-based readings, and are therefore not applicable
Sensors 2022, 22, 9878. https://doi.org/10.3390/s22249878 https://www.mdpi.com/journal/sensors
