AI-Based Modeling Software for Amine and Degradation Product Emissions

Significant commercial scale deployment of carbon capture technologies is necessary to achieve the climate change mitigation goal of greatly reducing global CO2 emissions while allowing for the utilization of fossil fuels as a transitional power source to continue. Carbon capture technologies that are currently being investigated and deployed throughout industry show significant promise in helping to mitigate the effects of climate change. Solvent-based carbon capture has been shown to be effective at removing pollutants from the flue gas. However, oxidative and thermal degradation of the solvents can lead to the formation of potentially hazardous chemical species. These additional species need to be carefully controlled, and where possible avoided, using best practices and engineering controls. Knowledge of proper solvent management techniques is critical to understanding how to mitigate the effects of these emissions over a full range of operating conditions. Machine learning (ML), a subset of Artificial Intelligence (AI), offers a powerful approach to analyzing large datasets and identifying patterns that can improve the accuracy of emissions predictions. Recently, a test campaign was carried out that measured emissions for post combustion capture from a residue fluid catalytic cracker. The data from this campaign can be used to train an AI based forecasting model that can predict emissions for solvent-based carbon capture systems. This tool could be used to help assess the impact of adding solvent-based to a specific process, improving the effectiveness of an existing solvent-based capture system, or integrated into a process control software application. The model will also be integrated with an advanced, hybrid energy decision-making software framework being developed by Reaction Engineering International (REI), which is funded by DOE Phase II SBIR. The ML model will make an important contribution to the resulting Phase II softwareĺs capabilities. The proposed effort will focus on the development and validation of a machine learning model for predicting emissions from plants with carbon capture technology based on amine-based solvents. The specific technical objectives for the Phase I research and development include the following: 1) Develop and validate a machine learning model for predicting emissions using historic test campaign data including operating parameters; 2) demonstrate the ability of the ML model to operate in a real-time environment within a plantĺs control system; 3) demonstrate the ML model as part of an advanced, hybrid power systems decision-making framework that REI is currently developing. The results of this research will have immediate application to currently operating solvent-based carbon capture systems and will help decision makers understand the impacts of future implementation of such systems. The proposed softwareĺs powerful simulation capabilities will be made available to end-users of amine-based capture systems that do not have dedicated modeling and simulation experts on staff. The modeling tools will provide real-time prediction of emissions products, the ability to investigate the effects of operating conditions on the emissions, and the ability to mitigate emissions using ôwhat ifö scenarios to seek optimal plant operational parameters