Application of the ABC Algorithm in Parameter Estimation and Kinetic Model Selection in Propionic Fermentation

A propionic acid fermentation process not only provides a more sustainable approach but also opens the door to propionic acid production capacity in regions with limited petroleum supplies. With fermentation, low-cost substrates can be used, such as residual biomass; reducing their concentration in nature. This process becomes interesting because from it propionic acid is considered natural. Several models have already been developed to describe the dynamics of components such as: Microorganism (biomass), nutrients (substrate), metabolites (product). However, a challenge is how to define the model that best represents the kinetic term, and therefore, there are several models for this modeling. This article's novelty is the application of the Bayesian technique (Computational Bayesian Approximation) to estimate parameters and simultaneously select the best model. Model validation was carried out considering propionic fermentation regarding experimental data from the literature, which selected the Andrews model as the best to predict the dynamic of biomass, substrate and product by the following parameters estimated = 0.192, ms = 0.005, mp = 0.017.