(27f) Dynamic Modeling with Uncertainty Quantification of Solid Sorbent Based CO2 Capture Processes

Post-combustion CO₂ capture is a potential approach to mitigate buildup of CO₂ concentration in the environment. Significant research is currently being carried out all over the world to reduce the penalty for CO₂ capture. Due to the large amount of CO₂ that needs to be captured, significant scale-up is needed, starting from the lab scale investigation of a potential capture technology to its commercial implementation for large-scale power plants. Since scale-up of novel chemical technologies generally extends over 20-30 years, it is imperative that the time required be significantly reduced for the accelerated commercialization of the potential technologies. Rigorous process models of these novel capture processes can be very useful for this purpose. However, process models suffer from uncertainty. The parameters for submodels are typically calibrated deterministically for each submodel. Further, the models are imperfect and the experimental data are not only imperfect, but are often inadequate. The resulting parametric and model-form uncertainties can lead to high discrepancy in the model outputs. Therefore, process models should be developed with quantified uncertainty, which can also be instrumental in identifying where additional experimental data should be collected and/or where different or modified model form should be considered to reduce the discrepancy.

With these motivations, a 1-D dynamic, non-isothermal and non-adiabatic fixed bed adsorption reactor model is developed for novel solid sorbent-based CO₂ capture processes. Rigorous models for mass transfer, heat transfer and chemical kinetics, if applicable, of both physisorption- as well as chemisorption-dominated processes are developed. A Bayesian calibration approach is used for quantifying parametric uncertainty. Model-form discrepancy is described by a set of orthogonal functions. Posterior distribution of the parameters as well as the model-form discrepancies are propagated through the process model to obtain the bounded confidence on the modelâ??s predictions of variables of interest such as bed pressure drop, breakthrough curve, and bed temperature profile. Our study shows that the output discrepancy strongly depends on the operating conditions and helps to identify approaches that can be used to reduce the scale-up uncertainty of the novel CO₂ capture processes.