AI tools fight ‘parasitic energy cost’ to enable carbon capture success

In the fight against climate change, we need to use all the tools at our disposal. Many policymakers are pinning their hopes on carbon capture being one of the most effective, but it is yet to make any serious impact on net zero efforts due to limited deployment, storage concerns, and technical challenges.

The most pressing of these is the “parasitic energy requirement”, according to Professor Peter Cummings, referring to the increased electricity demand caused by adding a carbon capture system to a fossil fuel power station. Recovering the solvent – which binds with carbon dioxide (CO₂) to prevent it reaching the atmosphere – creates a huge energy demand, said Professor Cummings to Professional Engineering. He hopes to tackle the problem using new AI tools.

Parasitic energy requirements mean power plants have to generate 30% more electricity, he said. “In principle you're generating more carbon dioxide, because the source of that energy to run the process is going to be from burning fossil fuels.”

Developing new solvents that have lower energy requirements is a key aim of the new ECO-AI project led by a team including Professor Cummings at Heriot-Watt University in Edinburgh. Announced this week (14 May), the scheme hopes to use AI to slash the development time for new carbon capture and storage (CCS) methods ‘from 100 days to 24 hours’, using the results to reduce the cost of the process.

“A lot of people, when they try to discover new solvents for carbon capture, they focus on one thing only – and that is the solubility of carbon dioxide in the solvent,” said Professor Cummings. “They don't look at the rest of the process – you have to pump the solvent that's got carbon dioxide in it over to the recovery column, where you get that carbon dioxide out. That requires heat, and then you have to then recycle the solvent.”

Researchers in Heriot-Watt’s iNetZ+ institute are developing neural network tools that can optimise the structure of solvent molecules to provide certain properties. “Given the structure of the molecule, given the conditions under which it's operating, we want to be able to predict the solubility of carbon dioxide in that fluid, what the viscosity of that fluid is, whether it's got carbon dioxide in it or not,” said Professor Cummings.

The work, in partnership with Imperial College London, will use process simulation to identify solvents that capture more CO₂ and use less energy over the entire process. Multiple AI models will be combined to generate “many more amines than are currently considered”, including some that have not yet been synthesized.

“We're calculating the properties of those using tried and true molecular modelling methodologies, molecular simulation and so on,” said Professor Cummings. “But those are very expensive calculations. Those are the ones that we're trying to replace with the AI models.”

The resulting tools could be 10- to 100-times faster than a fully physics and chemistry-based model, he said. “If we discover some better, cheaper, lower-energy consuming solvents for carbon capture, that would be a big deal… There's also societal outreach, for acceptance of carbon capture and sequestration. That has to be done.

“But I think the biggest win would be to come up with better solvents that will lower the parasitic energy costs and therefore lower the cost of doing carbon capture and sequestration.”

The research, which is funded by £2.5m from UK Research and Innovation, is also using AI tools to model CO₂ flow in geological storage, and forecasting levels of application in various industrial sectors. It aims to show that CCS could be “a viable economic option” for hard-to-decarbonise industries such as steel, cement and chemicals, a Heriot-Watt announcement said.

Finding geological storage has conventionally relied on complex, large-scale computational fluid dynamics to model the way CO₂ moves underground after injection, said Professor Cummings. Using AI models instead could speed up the process by “one or two orders of magnitude”, he claimed.

The ECO-AI project will last for two years, after which researchers hope to continue with additional funding.

Want the best engineering stories delivered straight to your inbox? The Professional Engineering newsletter gives you vital updates on the most cutting-edge engineering and exciting new job opportunities. To sign up, click here.

Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.