DESC0025123

The effective deployment of carbon capture technologies at industrial and electric power generation facilities is pivotal for reducing greenhouse gas (GHG) emissions and combating climate change. However, the integration of these technologies, particularly solvent-based carbon capture systems, introduces complexities related to emissions control and environmental management. Solvent-based systems, while beneficial in removing harmful pollutants like sulfur dioxide, face challenges due to the potential degradation of solvents. Traditional analysis methods offer limited insights and fail to capture the complex, multivariate behavior of these systems comprehensively. Impact Innovations LLC will develop a physics informed artificial intelligence (AI) based modeling tool to forecast amine and degradation product emissions from a carbon capture plant in real time. By leveraging advanced transformer-based time series models, the latest deep learning techniques for causal modeling, and generative AI for handling sparse data, the modeling tool will not only meet but also exceed current benchmarks in emission forecasting and mitigation strategies significantly. The expected cost savings and performance enhancements offer a compelling case for adopting this approach, aligning with the Infrastructure Investment and Jobs Act (IIJA)'s objectives to promote advanced carbon capture technologies and sustainable industrial practices. During Phase 1 of the project, Impact Innovations will (1) develop the proposed AI based modeling tool using data sourced from tests conducted at Technology Center Mongstad (TCM) with CESAR1 solvent, (2) assess and characterize the prediction accuracy of this tool, and (3) engage with industries operating carbon capture plants using amine-based solvents for further refining the toolĺs capabilities. Phase 2 of this project will focus on deploying an AI based software service. This service can be linked to the database of the carbon emission plants for real time data ingestion, training and visualizing the predicted results.