Innovative new machine learning techniques could help detect leaks during underground carbon dioxide (CO2) sequestration, protecting the environment while reducing unnecessary wastage.

A study being undertaken by researchers from Teesside University and international partners aims to explore the potential of machine learning and artificial intelligence (AI) to detect leaks during CO2 underground sequestration in pipelines and well string.  

Led by Dr Aziz Rahman, Associate Professor, Texas A&M University at Qatar $530,000 (£430,000) by Qatar Foundation Priority Research, the project is a response to the potential threat of CO2 leakage during storage.  

According to a study from Frontiers Energy Research, results show that carbon leakage can reduce the share of fossil based carbon capture and storage (CCS) by up to 35%, if it is controlled and correctly priced.  

Carbon leakage from CCS can lead to up to 25 GtCO2 (gigatonnes of CO2) of additional emissions throughout the 21st century for a leakage rate of 0.1% per year.  

The powers of AI can be harnessed by using Computational Fluid Dynamics (CFD) to predict both the likelihood and location of leaks in single-phase and multi-phase flow.  

To achieve this, the team will combine the machine learning approach with a ‘digital twin’ for leak detection during single phase (crude oil or gas) and multi-phase (multiple materials) flow during the transportation and injection of CO2 into the underground storage site.  

What is a digital twin?

A digital twin is essential a virtual object designed to emulate a physical object. By creating a virtual equivalent, researchers are able to run simulations and tests that can predict how a product or process will perform

Part of the process will see the creation of a virtual representation of a pipeline which is updated in real-time via a network of sensors mounted and installed in the real gas pipelines.  

Commenting on the project, Dr. Sina Rezaei-Gomari, Teesside University, said, “It is well-documented just how devastating leaks from pipelines can be if they aren’t spotted and acted upon in a timely and efficient manner.”  

“This research will look at how state of the art computational techniques including machine learning and digital twinning can be applied to accurately predict where faults are occurring.”  

By doing this, CCS companies and initiatives can use AI as a tool to prevent both small chronic and larger leaks, eliminating the need for human interference, saving both time and money.