Attention eLearning Users

We are upgrading our learning platform! As a result, if you have a course in progress, you'll need to complete it by December 24, 2021. Otherwise, you will need to restart the course beginning January 5, 2022 on our new platform. Repurchasing will not be necessary. Contact customer service with any questions.

(342m) Development of a Multi-Scale Model with Quantified Uncertainty for a Chemical Looping Process

Authors: 
Ostace, A. - Presenter, West Virginia University
Tong, A., Ohio State University
Mebane, D. S., National Energy Technology Laboratory
Burgard, A. P., National Energy Technology Laboratory
Miller, D. C., National Energy Technology Laboratory
Bhattacharyya, D., West Virginia University
Chen, Y. Y., The Ohio State University
Complex computational models are the basis of numerical simulations used to predict the behavior of systems and processes, to improve the safety and performance of existing technologies, and to aid the design and development of emerging, new technologies in an expedited timeframe [1,2]. Complex models are comprised of tightly coupled sub-models that represent physical and thermodynamic properties, reactions, mass and heat transfer, hydrodynamics, and other aspects of the process of interest. Uncertainty in the models describing the governing physics and chemistry of the system adversely affects the accuracy of the process models [3]. Uncertainty quantification (UQ) is widely recognized as an essential part of the numerical investigation and prediction of complex, non-linear system behavior [4]. UQ is of particular interest when process models are used for design of actual plants, where reducing the scaleup risk is of utmost importance.

Chemical looping combustion (CLC) is a novel approach to energy production with an inherent carbon capture capability, relying on metal oxides undergoing reduction-oxidation cycles to transfer oxygen from air to a fossil fuel between a fuel and an air reactor. In the fuel reactor, the fuel reacts with the oxygen carrier, generating an exhaust stream consisting of carbon dioxide (CO2) and water (H2O) from which H2O can be condensed resulting a CO2-rich stream ready for utilization or storage, while in the air reactor, the oxygen carrier is re-oxidized. By avoiding the direct combustion of fuel in air without relying on an expensive air separation unit, CLC is an attractive alternative to conventional CO2 capture technologies. However, CLC processes have not reached commercial realization yet, with widespread research efforts being focused on the development oxygen carriers, as well as on the design of suitable reactors.

Oxidation and reduction reactions in a number of potential oxygen carriers are highly complex and may involve various oxidation states that are difficult, if not impossible, to characterize and quantify as the reactions proceed. In addition, simultaneous diffusion and reaction of several gaseous species inside the evolving core of the oxygen carrier makes it difficult to develop models that can accurately describe the complex mechanisms. In this work, uncertainties in both model form and parameters for the kinetics of an iron-based oxygen carrier are quantified. Using a Bayesian framework, a kinetic model is calibrated to TGA data, quantifying the parameter uncertainty. The model-form uncertainty is quantified by using a Bayesian Smoothing Splines Analysis of Variance (BSS-ANOVA) model. This intrusive UQ enables the uncertainty in the form of joint model parameter and discrepancy posterior to be propagated into the process scale model.

The process-scale model used in this study is a 1-D steady-state, nonisothermal, first-principles model of a counter-current moving-bed (MB) reactor, applied to the simulation of the CLC of methane. Operating in the MB regime has the advantage of attaining high fuel conversion using relatively low solids circulation rates, improving overall process efficiency [5]. The MB model is implemented in the Institute for the Design of Advance Energy Systems’ (IDAES) equation oriented computational framework, which facilitates the development and solution of large-scale design and optimization problems. The Bayesian UQ framework yields probabilistic predictions of key process variables. Sensitivity to various operating conditions is evaluated providing valuable information on scale-up uncertainty.

[1] Miller DC, Agarwal DA, Tong C, Sun X, Tong C. CCSI and the role of advanced computing in accelerating the commercial deployment of carbon capture systems. SciDAC 2011 Conf 2011.

[2] Syamlal M, Guenther C, Cugini A, Ge W, Wang W, Yang N, et al. Computational science: Enabling technology development. Chem Eng Prog 2011;107:23–9.

[3] Li K, Mahapatra P, Bhat KS, Miller DC, Mebane DS. Multi-scale modeling of an amine sorbent fluidized bed adsorber with dynamic discrepancy reduced modeling. React Chem Eng 2017;2:550–60. doi:10.1039/C7RE00040E.

[4] Kennedy MC, O’Hagan A. Bayesian calibration of computer models. J R Stat Soc Ser B (Statistical Methodol 2001;63:425–64. doi:10.1111/1467-9868.00294.

[5] Okoli CO, Ostace A, Nadgouda S, Lee A, Tong A, Burgard AP, et al. A framework for the optimization of chemical looping combustion processes. Powder Technol 2020;365:149–62. doi:10.1016/j.powtec.2019.04.035.