Current overview and way forward for the use of machine learning in the field of petroleum gas hydrates

Abstract

Gas hydrates represent one of the main flow assurance challenges in the oil and gas industry as they can lead to plugging of pipelines and process equipment. In this paper we present a literature study performed to evaluate the current state of the use of machine learning methods within the field of gas hydrates with specific focus on the oil chemistry. A common analysis technique for crude oils is Fourier Transform Ion Cyclotron Resonance Mass Spectrometry (FT-ICR MS) which could be a good approach to achieving a better understanding of the chemical composition of hydrates, and the use of machine learning in the field of FT-ICR MS was therefore also examined. Several machine learning methods were identified as promising, their use in the literature was reviewed and a text analysis study was performed to identify the main topics within the publications. The literature search revealed that the publications on the combination of FT-ICR MS, machine learning and gas hydrates is limited to one. Most of the work on gas hydrates is related to thermodynamics, while FT-ICR MS is mostly used for chemical analysis of oils. However, with the combination of FT-ICR MS and machine learning to evaluate samples related to gas hydrates, it could be possible to improve the understanding of the composition of hydrates and thereby identify hydrate active compounds responsible for the differences between oils forming plugging hydrates and oils forming transportable hydrates.

Current overview and way forward for the use of machine learning in the field of petroleum gas hydrates