A hybrid model approach to estimate and predict the behaviour of the lignocellulosic fermentation process facing the synergistic effect of inhibitory compounds and mixed carbon sources

Abstract

The replacement of the oil energy matrix with one based on renewable energy is urgent. An example of this transition is the production of ethanol from lignocellulosic feedstocks through fermentation process. Lignocellulosic fermentation media has a high-level complexity, so it is common to integrate experimental observations and mathematical modeling to analyze the fermentation comprehensively. Currently, first-principle models have been applied to develop a predictive mathematical model. However, since microbial platform of interest is often only partially understood, mismatches often exist. A data-driven model can also be developed from experimental measurements, but this model has narrow applicability as it is tailored to describe input-output relationships contained in the training datasets, with no reliable extrapolation. Thus, we propose a hybrid modelling strategy to better predict a lignocellulosic fermentation preserving the generalizability and interpretability of the model. We hope that this can be achieved through mutual knowledge exchange, where the Brazilian biorefineries and natural resources background (Dr. de Morais) will be deeply enhanced by the UK know-how on modelling and optimization of bioprocesses (Dr. Shah). For this project, it was set four specific objectives (I) adjustment of first-principle model previously developed by the PhD. Candidate (research project linked FAPESP #2019/00103-0) (II) choose and development of the best artificial neural network (ANN) architecture for fermentation processes; (III) assembly of the ANN with the first-principle model to create a hybrid model; (IV) Evaluate the prediction, generalizability, and interpretability of the hybrid model. Therefore, we expect to apply a hybrid model approach to overcome the complexity and nonlinearity of lignocellulosic fermentation medias providing devices for future modifications in the microbial platform structure. (AU)