Model-based control methodologies for catalytic surface reactions

Kinetic Monte Carlo (kMC) simulations can accurately portray the dynamics of both the microscopic and macroscopic behavior of reactions on surfaces. Unfortunately, optimizing an input trajectory for real-time control based on kMC models is still computationally intractable. This paper discusses how kMC models of CO oxidation can be used as the basis for the design of three different MPC controllers that are implementable in real-time. These models are: (i) a mean field model that closes in the species coverages, (ii) a lookup table of linear state-space models generated from realizations of kMC simulations, (iii) a low-order model (checkerboard mean field) conceptualized from observation of kMC simulation microscopic states. Closed-loop simulations of a kMC plant with these three MPC controllers show that all the controllers can meet overall control objectives, but with varying levels of performance. The checkerboard mean field controller demonstrates the best performance.