This thesis develops a PDE-based modeling and real-time state estimation framework for an adiabatic packed-bed reactor (PBR) subject to thiophene poisoning. A pseudo-homogeneous model couples mass, energy, and deactivation balances to track reaction and thermal fronts as the catalyst loses activity. Numerical substepping and automatic differentiation handle the stiff PDE dynamics. A moving-horizon estimation (MHE) scheme reconstructs unmeasured states (a focus on catalyst activity) from limited temperature and composition measurements. Simulations confirm that an MHE equipped with a finite-difference representation of the PDE dynamics does not effectively track the poisoning front or predict the unmeasured variables. In the scenario where all state variables outside of catalyst activity are observable, the MHE effectively predicts catalyst deactivation in all stages of the PBR.
