James Nelson,* Robert Coleman, Christian Deakin, Becca Del Solar, Christine Benz, and Roger Boulton
*University of California, Davis, 1 Shields Ave, Davis, CA, 95616 (jjnel@ucdavis.edu)

The density-time curves of 32 commercial-scale wine fermentations from the 2021 harvest were analyzed using a fermentation model and several alternative methods of parameter estimation. The fermentations included both white and red fermentations that ranged in volume from 1200 to 420,000 L (320 to 110,000 gallons), across 17 dif­ferent yeast strains and at temperatures between 10 and 35°C (50 and 95°F). The es­timated model parameters included lag period, initial assimilable nitrogen concentra­tion, specific maintenance rate, cell viability constant, and ethanol inhibition constant. The methodologies investigated were the benchmark parameter estimation algorithm developed by Bard, a differential evolution approach, a particle swarm optimization, and a new seeded search technique applied to the Boulton fermentation model. The method efficiencies were compared and the model parameters used to classify the fermentations using multivariate methods. The results demonstrate the ability of these methods to analyze complete fermentation curves to obtain useful character­istics of fermentation performance at full scale and over a wide range of winemaking conditions. The methods also have applications to real-time fermentation monitoring and the prediction of peak energy transfer rates and fermentation completion..

Funding Support: This work was made possible by the financial support of the T.J. Rodgers Fellowship in Electrical and Computer Engineering, UC Davis (JN), Treasury Wine Estates (RC), E. & J. Gallo Winery, Healdsburg, CA (CD, BDS and CB) and the Stephen Sinclair Scott Endowment in Viticulture and Enology, UC Davis (RB).