Esterification of Levulinic Acid to Ethyl Levulinate over Amberlyst-15 in Flow: Systematic Kinetic Model Discrimination and Parameter Estimation

An automated reactor platform was developed using LabVIEW to conduct preplanned experiments for the identification of a kinetic model for the esterification of Levulinic acid (LA) and ethanol over heterogeneous Amberlyst-15 catalyst. A Single Pellet String Reactor of 1.25 aspect ratio was used for this kinetic study, loaded with 0.1 g of 800 μm catalyst spheres, at flow rates 20-60 μL/min, temperatures 70-100 °C, and LA feed concentrations 0.8-1.6 M. An extensive library of power law, Langmuir-Hinshelwood-Hougen-Watson and Eley-Rideal models, was screened through the application of a general procedure for model discrimination and parameter estimation. The procedure, consisting of seven steps, was applied for the investigation of different design spaces and allowed for the reformulation of models to include temperature-dependent parameters, the former leading to an increase in model identifiability and the latter resulting in enhanced model fitting. The combination of experimental data sets including the addition of the reaction product (water) in the reactor inlet stream and the incorporation of temperature dependence in the adsorption coefficients' expression led to the identification of two suitable kinetic models out of 28 candidates (a Langmuir-Hinshelwood-Hougen-Watson and an Eley-Rideal model), both of which accounted for the adsorption of water on Amberlyst-15 and fitted the experimental data satisfactorily.