A comparison between different modeling techniques for the production of bio-ethanol from dairy industry wastes

In the present work, the fermentation process aimed at obtaining bio-ethanol starting from ricotta cheese whey (RCW), a waste biomass rich in lactose, was simulated by both a pure neural network model (NM) and a multiple hybrid neural model (HNM). The simulation results showed that the developed HNM was capable of providing an accurate representation of the actual time evolution of lactose, ethanol and biomass concentrations even in conditions never exploited during model development. HNM predictions indeed exhibited an average percentage error lower than 10 %, as compared to the experimental data collected during RCW fermentation runs. The proposed methodology, leading to the formulation of a hybrid paradigm, may allow overcoming some of the inherent difficulties accompanying the development of reliable models that are called to describe the true behavior of biotechnological processes.