Towards Reliable CO₂ Capture: Nonlinear Optimization of a Piperazine-Based Process Under Uncertainty

<p dir="ltr">Meeting a net-zero greenhouse gas emissions target requires reducing CO2 emissions from fossil fuel based power plants and other industrial processes. Amine-based post-combustion capture processes have emerged as a prominent method to achieve this. Although many technological advancements have been introduced for such processes, leading to decreased energy requirements for capture, such as the Piperazine/Advanced Flash Stripper (PZ/AFS) process, there are still only a few commercial installations because of their high costs. Therefore, these processes can benefit from process optimization to enhance their economic viability. To that end, in the first part of the thesis, we present our equation-oriented model of the PZ/AFS process, implemented in the IDAES modeling framework. To ensure the robust convergence of this nonlinear model, we devise a multi-level cascade initialization scheme. We find optimal designs for different plant capacities, for two different flue gas sources, and for different target capture rates. </p><p dir="ltr">However, process models have many uncertain parameters that can make our cost estimates inaccurate and, more importantly, compromise the reliability of the deterministically optimal designs. Therefore, in the second part of the thesis, we identify the effect of economic un?certainties by a sensitivity analysis on utility and PZ solvent prices and evaluate different carbon tax scenarios. Then, focusing on epistemic uncertainties, we characterize and express the uncertainty associated with the parameters that have the greatest impact on cost as a 12- dimensional uncertainty set. Based on this, we solve a robust optimization problem to mitigate the risk posed by parametric uncertainties in our process designs. We then address operational uncertainties by solving a multi-scenario, two-stage robust optimization problem considering variations in feed flow rate and CO2 concentration. </p><p dir="ltr">Recognizing that solving nonlinear process models with complicating property models is challenging, in the third part of the thesis, we propose using an implicit function formula?tion for property calculations. We demonstrate this capability on a water desalination process optimization model.</p>