Modeling of Biomass Conversion to Furfural, A Platform Chemical: A Predictive Modeling Approach

It is crucial to predict the yield of platform chemicals from biomass to identify the commercial feasibility. However, yield greatly depends on biomass compositions and the reaction conditions. In this work, different biomass compositions and reaction conditions were successfully used as input to analyze the yield of one of the essential platform chemicals, furfural, using machine learning methods. Non-linear models could predict the yield of furfural better than linear models. Relative error analysis and scatter diagrams were used to analyze the predicted results. In addition, partial dependency analysis plots illustrate the influence of independent parameters on the target variables. This study provides feasible thinking for predicting the furfural yield obtained from biomass with different compositions under different reaction conditions.