Modelagem, simulação e discriminação estatística de modelos de um cultivo descontínuo de Saccharomyces cerevisiae utilizando o software livre EMSO

Abstract

Concern about environmental pollution caused by the continuous use of fossil fuels, especially oil, encourages the search for alternatives to minimize this problem. Ethanol stands out in this scenario because it is a renewable fuel and Brazil has an abundant source of raw materials. Ethanol can be obtained through the fermentation of sugars in the presence of the yeast Saccharomyces cerevisiae. The mathematical modeling of this process enables a better understanding of the physical-chemical phenomena involved and, consequently, assists in the design, control and optimization stages. Despite this, the vast majority of works available in the literature are purely experimental in nature. In this work, emphasis was placed on the use of the Environment for Modelling, Simulation and Optimization (EMSO), a free and opensource program, for modeling and simulating biotechnological processes. Experimental data from the literature regarding alcohol production with Saccharomyces cerevisiae was used in a batch bioreactor during 28 hours of fermentation. The kinetics were investigated using the Monod, Moser, Contois and Andrews models. The parameters of these models were estimated using non-linear programming (NLP) with a statistical confidence level of 95%. The criteria used to discriminate the models were the mean absolute deviation (MAD), the mean absolute percentage error (MAPE) and the square of the correlation coefficient between predicted and observed values (𝑅𝑦𝑐𝑦𝑒 ) 2 . Among the models investigated, the Contois model showed the best results, both for cell growth and substrate consumption. For cell growth, the MAD, MAPE and (𝑅𝑦𝑐𝑦𝑒 ) 2 values were 0.2451 g/L, 9.82% and 0.9737, respectively. For substrate consumption, the MAD, MAPE and (𝑅𝑦𝑐𝑦𝑒 ) 2 values were 4.1683 g/L, 17.67% and 0.9886, respectively.