Machine Learning Models for Absorption-Based Post-combustion Carbon Capture

Carbon dioxide has been identified as one of the leading causes of global climate change. For this reason, it is important to reduce carbon emissions associated with industrial activities. One method of achieving this goal is to replace older industrial processes with newer equivalent practices that produce less greenhouse gases. However, this is costly and will take a long time to implement. In addition, not all carbon heavy industrial processes have a suitable equivalent. Post-combustion carbon capture (PCCC) is a solution that can be implemented alongside existing infrastructure, such as steel mills and cement plants. The main barrier for widespread PCCC implementation is its large energy usage. Improving the energy efficiency of the carbon capture process may lead to greater adoption by industries. However, optimization using simulations requires an accurate model of the system. There are two main methods of developing models, mechanistic and empirical. Mechanistic models are built from the ground up using theoretical relationships between fundamental components of the system. Empirical models are based on retrospective observed and prospective experimental data. One subset of empirical models is machine learning, where an algorithm is used to identify relationships between a set of input and output variables. The first goal of this chapter is to provide an overview of machine learning concepts and general model architectures in the context of post-combustion carbon capture. The second goal of this chapter is to present and compare different machine learning models within the field of absorption-based carbon capture. The strengths and limitation of the strategies used in the creation of past models will be discussed.