Machine Learning in Carbon Capture : Utilization & Storage

Machine Learning in Carbon Capture : Utilization & Storage

This summer, I am participating in a research project at MIT Climate & Sustainability Consortium (MCSC). The project focuses specifically on topics related to carbon capture, utilization, and storage (CCUS) technologies and explores the methods to integrate the CCUS process, with a focus on approaches with commercial values, as well as building an efficient mechanism to upscale the most profitable solutions to add on to the existing infrastructure. Machine learning is applied at a molecular scale and process level, and such applications are very valuable in research in CCUS. 

For this project, I am taking a role focusing mainly on applying machine learning in carbon utilization.

I am using machine learning in direct air capture to dive into three fields:

predicting physical or thermodynamic properties, optimizing process conditions, and explicitly automating materials synthesis. 

Carbon utilization has several approaches, such as CO2-Enhanced oil recovery (CO2-EOR), CO2-Enhanced coalbed methane (CO2-ECBM), and the conversion of CO2 into valuable products such as chemicals, fuels, and building materials. Key benefits of using machine learning in such fields could be boosting computational speed, or easing the complexity of solving the problem, figuring out the unclear input-output pattern and structures that exist in the obtained simulated database, as well as predicting certain properties of products. I am focusing my research mainly in converting carbon into fuels. And the reason to use machine learning is that machine learning algorithms could efficiently screen the huge number of catalysts for the CO2 catalytic or electro-catalytic conversion.

There has been several noted contributions of valuable work done in this field. For example, Ulissi proposed to use a neural-network-based surrogate model together with density functional theory (DFT) calculations. As a result, enabling exhaustive searches for active bimetallic facets. And reveal active site motifs for carbon reduction. In addition, Zhong also claimed that through ML and DFT calculations. Moreover, Cu-Al electrocatalysts can efficiently convert CO2 to ethylene with the highest faradaic efficiency reported so far.

In conclusion, through learning the existing work as well as the different machine learning algorithms. Lastly, I am working on figuring out how to generate a perspective to identify new opportunities for machine learning in electrochemical best systems for carbon utilization.

Written by Nancy Cheng

Back To News

Physics Nobel Prize Winner MIT Prof Frank Wilczek. On String Theory, Gravitation, Newton & Big Bang

Machine Learning in Carbon Capture : Utilization & Storage