Jens Bremer

POD-DEIM for Efficient Reduction of a Dynamic 2D Catalytic Reactor Model

Abstract Many computational difficulties in dealing with chemical process models often result from spatially distributed states as well as nonlinear correlations (e.g., for transport coefficients or reaction kinetics). Surrogate models with sufficient accuracy represent one remedy to this problem. Featuring a lower number of states, model order reduction (MOR) generates considerably less complex models and leads to faster model evaluations. Especially for nonlinear systems, snapshot-based MOR techniques are considered to be one of the most promising methods. In this study, we apply proper orthogonal decomposition together with the discrete empirical interpolation method (POD-DEIM) to a dynamic, two-dimensional reactor model for catalytic carbon dioxide methanation. Motivated by renewable energy integration, we consider this reactor in two different dynamic scenarios: Disturbed continuous operation and start-up. It can be shown that the reduced order model (ROM) is accurate and, furthermore, the solution of the FOM is accelerated at least by one order of magnitude.

Nonlinear Model Order Reduction for Catalytic Tubular Reactors

Abstract Dealing with dynamic, nonlinear, large-scale process models often leads to many computational difficulties. One remedy to this problem is the replacement of these large-scale models by surrogate models with sufficient accuracy. Model order reduction (MOR) generates considerably less complex models featuring a lower number of states, leading to faster model evaluations. Snapshot-based MOR techniques are considered to be one of the most promising methods for nonlinear systems due to their flexibility and universal applicability. In this study, we apply a widely used MOR technique, namely proper orthogonal decomposition together with the discrete empirical interpolation method (POD-DEIM), to a dynamic, two-dimensional reactor model for catalytic H 2 -methanation. As simulation application, we consider the start-up behavior of the reactor when operated with hydrogen generated by use of volatile renewable energy. It can be shown that the reduced order model (ROM) is accurate and, furthermore, the solution of the ROM is accelerated at least by a factor of 125.

Physics-Based Surrogate Models for Optimal Control of a CO2 Methanation Reactor

Abstract Transition to renewable energy sources requires energy storing technologies like Power-to-Gas approaches to counter their inherent volatile nature. As a consequence, dynamic reactor operation becomes necessary. The exothermic carbon dioxide methanation is one key conversion step within the Power-to-Gas process, which requires careful reactor temperature control due to distinct hot spot formation. The successful yet computational expensive identification of a time optimal reactor start-up control trajectory by Bremer et al. (2017b) demonstrates the need for computationally efficient but accurate surrogate modeling approaches with regard to online applications (e.g., NMPC). In this study, two physics-based surrogate models, a fully one-dimensional and an extended Alpha model approach (Hagan et al., 1988), are compared to the two-dimensional model in terms of accuracy and computational load. The results indicate a limited applicability in terms of hot spot temperature prediction and the need for further investigation of efficient surrogate models for dynamic operation of highly exothermic reactors.