(137c) Hydrotreating of Bio-Oil Model Compounds: Experimental Results and Kinetic Modeling

Introduction

Although bio-oil from fast pyrolysis of lignocellulosic biomass is a potential source for fuels and valuable chemical building blocks, such bio-oils cannot be directly integrated into existing petroleum refining facilities. Indeed, fast pyrolysis bio-oil has an extremely complex organic character with not only a high amount of carbon, but also large quantities of oxygenated species. Since the presence of oxygen reduces the quality of bio-oils as a fuel source, bio-oils require a dedicated pre-refining step, which is generally a hydrotreating process. However, due to its high complexity, in both molecular and reaction terms, the upgrading of bio-oil reveals to be difficult. To correctly design such a process, one must acquire insight in the molecular composition of bio-oils and of the reactivity of the various species.

The present work describes a methodology for modeling the upgrading of bio-oils. The proposed method will be applied to construct kinetic models for the hydrotreating of bio-oil model compounds. This strategy will enable a further understanding of bio-oil reactivity, opening the doors for the modeling techniques for full-range bio-oils.

Methodology

As modeling strategy, a direct discrete simulation of reaction events was chosen. The approach is based on a kinetic Monte Carlo (kMC) method, termed Stochastic Simulation Algorithm (SSA)^1,2, and describes the evolution of a system, molecule by molecule. Unlike the classic procedures, this method can describe the evolution of a molecular mixture over time without a pre-defined reaction network.

The stochastic approach is based on the identification of all possible reaction events at a given moment in time and on the calculation of the reactivities for each event. Thanks to the work of Gillespie^1,2 and recalculating apparent first order rate coefficients for the reactions, the reaction probability can be generated for each reaction event. By assembling all the probabilities of all molecules, the overall reactivity allows to determine, via Monte Carlo sampling, the time interval until the next reaction, and to select a given reaction. The sampling process of both variables is then repeated until the final simulation time is reached.

The SSA requires as inputs a list of reactant molecules, the different types of reactions that can occur and the kinetic parameters associated to these transformations. Due to its stochastic nature, this method generates a different outcome for each simulation, which does not allow to construct the general evolution of the reaction system over time. Therefore, as with any stochastic approach, the method has to be based on several simulations, each of which calculates discrete time trajectories of molecular populations. Their average profile will converge to the same results as deterministic methods.

Results

To understand the reactivity, the hydrotreating of two model molecules was studied: Guaiacol and Furfural. The experimental data was obtained by Ozagac³ and by Costa da Cruz⁴ for the hydrotreating of guaiacol and furfural over a NiMo/Al₂O₃ catalyst in a batch reactor at 13 MPa, at 200°C, 250°C, and 300°C, and for reaction times of 3h and of 5h. For several experimental conditions, blank runs were performed and showed a relatively high conversion. For guaiacol, an initial conversion of 10% was obtained at 300°C after the heating up and cooling down the batch reactor.

For the kinetic modeling of hydrotreating guaiacol and furfural, the reactions for hydrogenation, dehydrogenation, demethylation, demethoxylation, decarbonylation, decarboxylation, deoxygenation, transalkylation, and dehydration were implemented in the Stochastic Simulation Algoithm. For the model molecules, the SSA method is able to predict the expected trends, not only for the conversion of the reactants, but also for the generation of the various products and for the hydrogen consumption. Additionally, this approach also predicts the transformation of the several reaction products into smaller molecules. Due to the high reactivity of guaiacol, the heating and cooling periods contribute significantly to the overall conversion. Hence, the reactor simulations need to account for these dynamic periods.

The model results were compared to the experimental data. Accounting for the thermal behavior of the reactor during heating, reaction and cooling, the kinetic parameters of the reactions were estimated. A good agreement is observed. The comparison of the simulated and the experimental selectivities for methanol and dialcohols shows that the model predicts, within the error of the experimental data, the correct order of magnitude for both chemical families.

Conclusions

The SSA methodology was applied to bio-oil model molecules and showed the correct trends when compared to available experimental data. Through this study, a kinetic model was generated that is able to simulate the reactions of guaiacol and furfural under hydrotreating conditions. It was shown that the SSA method is an excellent simulation tool for complex reaction networks without the need for a pre-defined reaction network. Furthermore, this method enables a better understanding of these systems, making it suitable to be applied to a full range bio-oil.

Literature Cited

1. Gillespie D.T., “A general method for numerically simulating the stochastic time evolution of coupled chemical reactions,” Journal of Computational Physics, 22(4), pp. 403–434 (1976).
2. Gillespie D.T., “A rigorous derivation of the chemical master equation”. Physica A: Statistical Mechanics and its Applications 188(1-3), pp.404–425 (1992).
3. Ozagac M., “Etude mécanistique de l’hydroconversion catalytique de bio-huiles de pyrolyse,” PhD thesis, Université Claude Bernard - Lyon 1 (2016).
4. Costa da Cruz A.R., “Compositional and kinetic modeling of bio-oil from fast pyrolysis from lignocellulosic biomass”, PhD thesis, Université Claude Bernard - Lyon 1 (2019).