advanced deep regression models for smart operation of the oil and gas industry
abstract
the first industrial revolution in the early 18th century largely exploited steam power to replace animal labor. since then, there has been rapid development in industrial operations. now, the world
has come to the brink of the fifth industrial revolution, a.k.a. industry 5.0, where industries invest in
building intelligent systems to perform complex actions more efficiently by leveraging technological advancements, including big data, and high-performance computing (hpc) platforms. thus,
modern artificial intelligence (ai), particularly deep neural networks (dnns) has emerged as a
powerful tool in industries for informed operational control, real-time fault and anomaly detection,
and maintenance.
in this direction, this research aims to develop advanced regression models using artificial
neural network (ann), 1-d convolutional neural network (cnn), and long short-term memory
(lstm) units for key operations in the oil and gas industries. more specifically, this study focuses
on three stages, namely drilling, transportation, and production, and proposes robust regressors for
accurate prediction of void fraction, the temperature of internal components of electric motors, and
the production level of hydrocarbon extracts. a precise prediction of these factors will increase
resource efficiency, energy saving, and product quality, and decrease environmental pollution. an
exhaustive experimental study conducted on benchmark datasets demonstrates the practicability
of the proposed solutions and their robustness. it is worth mentioning that canada is the world’s
fifth-largest oil producer and has one of the world’s largest oil reserves and the world’s third-largest
proven oil reserves.