How Hybrid Neural Model Might Help on the Optimization of Biofuels Product

Abstract

The aim of this paper was showing how artificial neural networks could by applied in modeling and optimization of biofuel production processes. In bioprocesses the use of biocatalysts, like enzymes or whole cells, usually requires a detailed analysis of kinetics and mass transfer phenomena difficult to be achieved using a pure theoretical approach. Artificial neural networks (ANNs) are black­box models that may be used in bioprocess modeling both realizing pure empirical models and hybrid neural models (HNMs), i.e. a combination between pure black­box and theoretical models. In this paper three case­studies were reported referring to biodiesel, bioethanol and biogas production starting from agro­industrial wastes. When a partial knowledge of the process at hand was available, a hybrid neural approach was applied (case­study 1­2), whereas when the complexity of the analyzed phenomena did not allow any reliable theoretical analysis, a pure black­box model was developed (case­study 3). In case­study 3, the possibility to incorporate the realized model in an optimization procedure was shown as well. The obtained results testify the effectiveness of neural networks in bioprocess modeling and optimization and suggested the exploitation of ANNs that represent a very flexible tools to describe bioconversion and bioreactors systems.