A Hierarchical Multiscale Approach for Predictive Microkinetic Modeling of Hydrogen Production

An important objective in process development is optimal reactor and operating conditions design. In order to accomplish this, one needs reliable, predictive, and efficient chemistry models. Development of such chemistry models is a difficult task for complex processes; the main problems include lack of kinetic parameters, reaction pathways, thermodynamic consistency, insufficient validation, and pressure and materials gaps. Therefore, traditional modeling approaches, including microkinetic analysis, have only partially succeeded in developing predictive models. This research work has two main objectives: 1) to propose a novel hierarchical multiscale approach for development of predictive microkinetic models to overcome the current limitations, and 2) to apply this approach to the development of microkinetic models for various hydrogen generation processes.

To achieve the first objective, we focus on the following features: 1) Bottom up mechanism building: This feature indicates a sequential mechanism development from simple to more complex systems. 2) Hybrid parameter estimation: Parameters are estimated using a variety of semi-empirical tools, literature information from surface science and reactor experiments, and first principles techniques. Parameter estimation on a number of catalyst surfaces helps to overcome to some extent the materials gap problem. 3) Microreactor experiments: In case of lack of certain experimental data, we have conducted those experiments in the laboratory. For example, experiments for H2 ignition in a microreactor have been useful for the development and optimization of a H2 oxidation microkinetic model. 4) Thermodynamic consistency: Thermodynamic inconsistency is a long-standing problem resulting in incorrect rate calculations, energy balance, and equilibrium predictions. A combination of semi-empirical methods and statistical mechanics with constraints-based optimization is proposed to ensure thermodynamic consistency in catalytic reaction mechanisms. 5) Important parameter identification and iterative refinement: Rather than using the computationally expensive first principles techniques for estimation of all parameters, we propose that first principles methods are used on-demand only for the refinement of important parameters in a mechanism. Important parameters are identified using sensitivity analysis, coverage analysis, etc., and only those parameters are refined using first principles techniques or optimization. This particular feature results in tremendous CPU savings. 6) Mechanism validation and analysis: Sensitivity analysis and diversity in operating conditions are employed to classify available experimental data as target or redundant types. The mechanism parameters are developed and refined using the target experiments, but a number of redundant experiments are also employed to validate a reaction mechanism over a wide range of operating conditions, which helps to overcome to some extent the pressure gap. To understand the underlying physics of a process, the overall flow of reactants to products, and the rate-determining step for a particular model response, reaction path analysis is employed. 7) Computer-assisted model reduction: With a large microkinetic model, the bottleneck in reactor design using computational fluid dynamics is the computational power. Therefore, simple one-step rate expressions are developed for efficient implementation in reactor design.

As a second objective, we apply this approach to the development of predictive microkinetic models for H2 production. H2 production from fuel processing is a crucial component in the blooming fuel cell technology. A number of catalytic processes such as CH4 partial oxidation, steam reforming, dry reforming, autothermal reforming, etc., could be employed for H2 generation, followed by water-gas shift and preferential CO oxidation for CO cleanup. NH3 decomposition is also an attractive alternative. In this work, we have developed a total of 12 microkinetic and 2 reduced models on 3 catalysts, for these H2 production processes.