Largely due to CO2 emissions, the earth is 1.1℃ hotter than it was before industrialization with a very real chance of irreversible temperature increases and damage. What if there was a way to reverse these destructive CO2 emissions? The answer may lie in carbon capture and storage systems (CCS), both biological and mechanical.
Since the first oil well was dug in 1859, we have been extracting carbon from the ground and pumping it into our atmosphere.
Similar to the hundreds of thousands of oil wells around the world, what if we had many CCS systems capturing carbon, often in the form of oil, and injecting it back into the ground.
In this article we highlight organizations engineering mechanical solutions that remove CO2 from the atmosphere. While mechanical CCS shows great potential to help reduce CO2 emissions, Artificial Intelligence (AI) is proving instrumental in achieving large scale deployment of CCS which in turn might help lead to a long term greenhouse gas (GHG) emissions solution.
Mechanical CCS solutions have received both criticism and applause. One of the major reasons people are skeptical about the viability of this technology is the high energy requirement of direct air capture (DAC) systems, a type of CCS. Last September, Climeworks opened the first and largest direct air capture and storage plant in Iceland. The direct air capture (DAC) system works like a fan bringing in outside air. Air moves through a filter where carbon dioxide is trapped inside. Once the filter is full, it is heated to release CO₂. The air captured CO₂ can be upcycled as a raw material or stored for near permanent removal. In Iceland, via the Carbfix technology, it is mixed with water and then pumped underground where it mineralizes for safe storage in under two years. (1) DAC systems use a lot of energy in this process and because of this, many large scale DAC systems turn to fossil fuels to get enough energy to power themselves; (2) this fossil fuel usage seems counterintuitive and is often cited as a criticism.
Photo: www.climeworks.com
An example of how AI can be used to automate a process in DAC systems is through new material development. DAC systems use adsorbents to separate carbon molecules from the other thousands of molecules in our atmosphere. Scientists are discovering new materials that can make this process cheaper and more efficient. In the past, scientists were often in a lab physically testing chemicals trying to discover new materials. Now with the advancements of AI, this process can be done without the expense of manual labor. ML models can be trained to find optimal materials with low vapor pressure and low thermal degradation components to enhance the performance of DAC systems.
TotalEnergies is a company currently using AI for new material discovery. They use these newly discovered materials to optimize their process for capturing carbon from the atmosphere. TotalEnergies uses AI technology to identify new materials that will effectively capture carbon through Direct Air Capture and Combined Cycle Gas Turbines. (3) TotalEnergies is also collaborating with Shell and Equinor to build a long-term carbon storage facility in Norway. This project has the potential to reach a storage capacity of 1.5 million tons of carbon per year. Also, because of the large storage capacity, this project has the potential to receive CO2 from multiple companies around the world. (4)
“At TotalEnergies, we use AI to discover and design new materials for capturing CO2, especially for DAC (Direct Air Capture) and CCGT (Combined Cycle gas Turbines). AI applies to optimize the CO2 capture process accordingly. We also consider AI for history matching of CO2 storage simulation. These are the ways our teams use AI today to achieve our ambition to become net zero emission together with the society.”
Machine learning (ML) algorithms are used by companies like Google to track energy consumption and identify areas to increase efficiency. Google has seen a 40% decrease in energy used for a data center cooling system since this process started. Large AI models are often criticized for high energy usage. Training can indeed use a lot of energy, however, typically the energy and cost savings at inference time outweighs the energy used to train the model. (5)
“IBM’s team
IBM’s team has been using AI to quickly determine which types of rock are best for storing carbon. They have been using a combination of high-performance computing (HPC) and hybrid cloud to build a new algorithm that can approximate tiny empty spaces, called capillary networks, in porous rocks that naturally occur in geological formations. These networks offer significant potential to store carbon dioxide – in either a liquid or solid form – that have been captured from flue gas or other sources at the point of emission. According to IBM, “Our algorithm performs a core analysis of a rock formation faster and more accurately than using lab tests alone. Our results could help cut the time for completing rock analysis from months to days, drive down costs, potentially increase efficiency, and reduce risks of geological carbon storage.” (6) When the carbon is captured and pumped into the earth, it is in the form of a supercritical fluid. Supercritical fluids are achieved through intense heat or pressure applied to a substance which creates a fluid with the combined properties of liquid and gas. (7) In this case, the further down you travel into the earth’s surface, the higher the temperature and pressure is which creates this supercritical fluid property. Finding rock that is most suitable for the carbon to be stored is critical to ensure the greatest quantity of carbon storage.
While the carbon fluid is expected to flow throughout the rock, there are different safeguards in place to prevent any sort of leakage. There are multiple techniques used to properly contain these carbon deposits such as trapping the carbon within a layer of seal rock from which carbon cannot escape. This way, once the carbon is captured, it can be stored for long periods of time without risk of leakage or damage to the vessel which might cause harm to the surrounding environment.
Photo: DOE.gov
In a similar way, Halliburton Landmark has done significant work creating a carbon storage model with the help of ML. In this case, storage capacity volume is estimated using AI. According to Halliburton Landmark, “To achieve realistic estimates of carbon capacity, various sources of relevant data is integrated and ingested. This data is then subjected to analysis, pre-processing and data engineering using numerous derived seismic attributes to augment additional features. This enables us to compute a Relative Storage Index (RSI) through a deep learning-based ML algorithm.” The system uses data to predict the size and storage potential of underground geological formations. These geological models are used to predict storage capacity on a sub-surface level. (8) Anyone looking to find the best location for underground carbon storage might benefit from this machine learning solution.
“Using AI for climate control could help reduce 2.6-5.3 gigatons of GHG emissions and could provide $1-3 trillion in value added when applied to corporate sustainability generally.”
CCS systems have been a controversial topic for years now, often seen as a distraction from the very necessary work of reducing emissions. We are, however, rapidly approaching a point of no return, and direct approaches not mired in politics might prove a prudent hedge. With the adoption of AI, some of the shortcomings of CCS have the chance to be avoided in the areas of time, scale, and energy as discussed in this article. Hopefully this article has shed some light on a few examples of successfully employed carbon capture systems leveraging AI.
Sources
climeworks.com
theconversation.com/why-the-oil-industrys-pivot-to-carbon-capture-and-storage-while-it-keeps-on-drilling-isnt-a-climate-change-solution-171791
totalenergies.com/news/totalenergies-initiatives-promote-ccus?folder=9186
totalenergies.com/media/news/press-releases/statoil-shell-and-total-enter-co2-storage-partnership
www.scu.edu/environmental-ethics/resources/ai-and-the-ethics-of-energy-efficiency
medium.com/mytake/artificial-intelligence-explained-in-simple-english-part-1-2-1b28c1f762cf
www.scimed.co.uk/education/what-is-a-supercritical-fluid
www.landmark.solutions/Spotlights/ID/43/Accelerate-Carbon-Capture-Storage-Innovations-with-Artificial-Intelligence