Résumé
Alcoholic fermentation is an important step of wine making process. Automatic techniques using on line monitoring of fermentation, modelling and process optimisation would be a valuable new instrument for winemakers. In fact, the fermentation duration and the total energy required for temperature regulation could be predicted and optimised, not only in one tank but also in the whole winery. We proposed a dynamic model of alcoholic fermentation based on the main yeast physiological mechanisms. This model consists of ordinary differential equations including numerous parameters that need to be identified and important interactions between explanatory variables. The model predicted accurately the fermentation kinetics of >80% of a large number of experiments performed with 20 wine yeast strains, 69 musts and different fermentation conditions.Thanks to the wide domain of validity of the model, a simulator based on this model was developed to help winemakers to optimize tank management. It not only predicts the end of the fermentation and changes in the rate of fermentation, but also includes an optimization module based on fuzzy logic. Optimized temperature profiles and nitrogen addition strategies are proposed to decrease the duration of fermentation and energy requirements at winery scale, according to user specifications.