(591a) Dynamic Timescale-Based Reduced Order Models for Simulating Moving Bed Chemical Looping Combustion Reactors

As the energy industry seeks to reduce its carbon footprint and operate in a dynamic energy ecosystem, simulations will become increasingly important to evaluate and screen alternatives for carbon-based energy production. One such alternative is the chemical looping combustion (CLC) process, in which oxygen is transferred from air to fuel through an oxygen carrier that continually loops between oxidation and reduction reactions occurring in separate reactor beds. The gaseous product of the reduction bed is then concentrated CO₂ and water vapor which can be easily separated for carbon capture or sequestration.

A CLC reactor is modeled as a system of partial differential and algebraic equations, which, when discretized into a large-scale system of differential algebraic equations (DAE), is challenging to solve numerically. For online applications such as control and real-time optimization, fast and accurate reduced order models (ROMs) are needed. In many chemical engineering applications, dynamic models exhibit timescale multiplicity that can be exploited in the creation of such ROMs. This is done by identifying states with fast dynamics and applying to them a quasi-steady state approximation. The challenge is to identify which states have fast dynamics when timescales are not explicitly apparent.

In this work we present a dynamic first principles-based PDAE model for a moving bed chemical looping combustion (MBCLC) reduction reactor as a part of the Institute for the Design of Advanced Energy Systems (IDAES) process systems engineering framework. Reduced order models are then constructed using timescale-based model reduction techniques and are evaluated in terms of accuracy and computational expense for relevant step changes in operating conditions. The MBCLC model simulates the oxidation of methane and reduction of an iron-based oxygen carrier in a counter-current gas-solid tubular reactor. Differential equations model are mass and energy balances, while algebraic equations define thermodynamic properties, transfer coefficients, and reaction rate. Reduced order models are constructed by applying singular perturbation and singular value decomposition methods to identify fast, stable states and construct a ROM valid over the state-space operating region. Combining the reduced order model with a suitable orthogonal collocation spacial discretization strategy, we develop a fully discretized ROM that is suitable for online model-based optimal control.