Carbon capture and storage currently costs too much and moves too slowly for widespread use. This project uses artificial intelligence to tackle three specific barriers: the energy-intensive chemical process of capturing CO₂, the high cost of finding safe underground storage sites, and the financial uncertainty that blocks investment in CCS projects. The research has three strands. First, AI will replace expensive molecular simulations with faster machine-learning tools to discover new liquid solvents that dissolve CO₂ more efficiently. Second, AI-powered "digital twins" will model how injected CO₂ moves through underground rock formations, linking small-scale flow behaviour to large-scale patterns while accounting for geological uncertainty—something standard simulations fail to do. Third, network science and forecasting methods will model how innovation rates, government policy, and carbon markets interact to influence CCS adoption. If successful, the work could lower the cost of CCS across the entire chain—from capture to storage to financing—making it a viable tool for reducing industrial CO₂ emissions at scale. The public would see no direct change, but the power plants, steel mills, and cement factories that supply modern infrastructure would have a clearer path to decarbonisation.
View original technical description
The International Energy Agency (IEA) has identified Carbon Capture and Storage (CCS) in deep geological formation as one of the key approaches to reduce CO2 emissions. CCS is a combination of technologies for CO2 capture from large emitter industries and CO2 storage in deep geological formations, preventing its release back into the atmosphere. Currently, there are key barriers for the wide adoption of CCS on a large scale, such as (a) the high cost of CO2 capture that is an energy intensive chemical process, (b) the high cost of subsurface CO2 storage especially at the early stages of site selection and characterization of safe storage sites, and (c) the uncertainties in how CCS projects are financed and the interplay between technological innovation and policy intervention on the CO2 market and emission targets. In this project, we aim to utilize our expertise in AI to address these barriers. The first is to accelerate material discovery for energy efficient CO2 capture using liquid solvent (a type of liquid that serves to dissolve CO2). In this task, AI aims to replace standard expensive predictive methods (using molecular dynamic simulations) with fast and robust tools using machine learning. Further, the search of possible solvents will be accelerated by using effective tools developed by the AI community for high dimensional optimisation and control. For the CO2 storage site selection, numerical simulations provide a pathway to understand the long-term fate of injected CO2 and risks of leakage back to the atmosphere. However, standard numerical simulations are expensive, fail to propagate flow information from the small-scale to the large-scale flow features and generally underestimates the geological uncertainty. In this task, AI will be used to model flow in the subsurface by fast digital twins to help design and manage CO2 storage with an ability to link scales and include all sources of uncertainty. Recently, we have developed a new, and potentially revolutionary, AI methods using repurposed AI software libraries to implement some of the standard numerical methods applied in computational physics codes to gain platform-independent codes with increased performance. Further, AI libraries are much easier to couple and allows us to bridge information across-scales effectively. Financing CCS projects necessitate policy intervention. We employ network sciences and novel forecasting methods to study and understand the complex interaction of the rate of innovation, policy and CO2 markets on adoption of CCS technologies. In summary, we will develop AI techniques to decrease the cost of CCS projects via advance simulation techniques, better financial modelling and discovery of new energy efficient capture solvents.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know