Développement de modèles cinétiques multifactoriels basés sur des relations structure-réactivité et différents modes thermiques pour la production de molécules plateformes à partir de sucre lignocellulosique

A transition to sustainable and renewable green energy sources is needed to decrease the extensive dependency on fossil-based products and mitigate their harmful environmental impacts. The sugars derived from lignocellulosic biomass (LCB) can be transformed into valuable platform molecules such as 5-hydroxymethyl furfural (5-HMF), 5-alkoxymethyl furfural (RMF), alkyl levulinates (RL), and GVL, through chemical conversion methods. Similar to an oil refinery, the concept of an integrated biorefinery aims to exploit the LCB potential fully. However, due to the overabundance of reaction pathways and target products, the current approaches are focused on a single product, which is not sufficient to cope with today’s economic demand. Additionally, the thermal risk must be addressed to ensure safe and optimal operating conditions. Considering these challenges as motivation, this research project aims to develop novel multifactor kinetic models capable of optimizing multiproduct processes for the transformation of sugars to platform molecules. In the first part, the application of the structure-reactivity concept was used to study the solvolysis of 5-HMF to RL in different alcohols. Linear free energy relationships (LFERs) were employed to account for solvent and substituent effects, allowing for the representation of the reaction system within a single kinetic model. In the second part, an assessment of the kinetics and energy density of these platform molecules from the solvolysis of fructose was conducted. In this study, we seek to evaluate the kinetic barriers that could appear when producing 5-alkoxymethyl furfural (RMF) and alkyl levulinates (RL), and how this limitation can affect the solvent selection; from this study, we can identify the best solvents for 5-RMF and RL, considering their energy density as a property of interest. Finally, the assessment of kinetic models in different thermal modes (isothermal and adiabatic) was performed using the tree structure Parzen (TPE) optimization method and cross-validation, which is essential for thermal risk assessment. From the structure-reactivity, the results obtained showed that the combination of the Taft and KAT equations for the substituent and solvent can be used to predict the kinetics of different levulinates; experimentally, it was found that their production rates follow the order: r_ML^MeOH>r_EL^EtOH>r_PL^PrOH>r_BL^BuOH>r_PeL^PeOH>r_HeL^HeOH>r_LA^H2O. The kinetics and energy assessment showed that ethanol is the most convenient for 5-RMF and RL production under the operating conditions, while long-chain alcohols are better for 5-RMF production. The thermal-based modeling demonstrates that TPE can be used to estimate kinetic constants efficiently and that stratified cross-validation can effectively evaluate and select models with similar complexity across both isothermal and adiabatic data. Furthermore, different ratios of these thermal mode data provide the means to assess the consistency in prediction capacity across various scenarios.