(204e) Model Predictive Control of an Experimental Protonic Membrane Steam Methane Reforming System

Optimizing the thermo-electrochemical process of generating and purifying hydrogen in protonic membrane reforming systems can facilitate the commercialization of this emerging reforming technology [1]. However, within the multilayered and cascading chemical interfaces of these systems, there are challenges governing the nonlinear dependence of thermal hydrogen production on electrochemical hydrogen separation. Stochastic factors such as catalytic activity and membrane conductivity also directly influence the maximum production rate of pure hydrogen and the energy efficiency of the reformer unit operation. Advanced control methodologies may regulate these factors by way of observer-based model predictive control architectures that improve the longevity of the reformer materials while maximizing the purified hydrogen target product, minimizing computational solution times, and updating controller models in real time.

In view of this, we have designed and implemented the first decentralized model predictive control (DMPC) architecture for the steam-to-carbon ratio of reactants, the hydrogen purification rate, and the overall rate of hydrogen generation in a 500 W protonic membrane reforming system. The transient responses of these key process variables were captured in validated-models derived in Cui et al. [2], and were subsequently implemented into the three predictive controllers studied in this work. An offset-free disturbance observer was also used to mitigate plant-model mismatch in the hydrogen generation rate control loop by adapting the reaction engineering model to shifts in catalytic activity, cell voltage spikes, local dehydration of membrane surfaces, or sensor drift.

Compared to a classical multi-input multi-output control scheme, the predictive controllers of the DMPC architecture show an improved time-to-setpoint, an improved adherence to process-specific constraints, and enhanced hydrogen purification rates exceeding 65%. Further, automated setpoint dynamic compensation of the hydrogen purification rate prevents the complete dehydrogenation of the protonic membrane, thereby limiting carbon formation reactions. Near-complete methane conversion at reformer temperatures below 800 °C is also observed. Generally, these experiments contribute to scale-up science [3] and reaction engineering [4] by quantifying the emergent phenomena in dynamically-operated thermo-electrochemical systems to provide potential automation pathways and enable optimal control.

[1] Fjeld, H.; Clark, D.; Yuste-Tirados, I.; Zanón, R.; Catalán Martínez, D.; Beeaff, D.; Hernández Morejudo, S.; Vestre, P.; Norby, T.; Haugsrud, R.; Serra, J.; Kjølseth, C. Nature Energy 2017, 2, 923–931.
[2] Cui, X.; Peters, D.; Wang, Y.; Çıtmacı, B.; Richard, D.; Morales-Guio, C. G.; Christofides, P. D. Chemical Engineering Research and Design 2024, 212, 493–519.
[3] Moore, T. et al. Nature Chemical Engineering 2024, 1, 731–740.
[4] Luterbacher, J.; Weckhuysen, B.; Haussener, S.; Cuenya, B. R.; Resasco, D. E.; Morales-Guio, C. G.; Jiao, F.; Zheng, N.; Domen, K.; Concepción, P., et al. Nature Chemical Engineering 2025, 2, 156–159.