DigitWine Project

Keywords: digital twins, wine fermentation, energy footprint, aromatic quality

Project Description

Context

World production of wine is around 25 billion litres per year and France is the world's second largest producer (17%). The wine industry is undergoing major transformations in response to global climate change and societal concerns. In particular, it needs to reduce its energy footprint, while maintaining the organoleptic quality of wines. 

Although winemaking fermentation is naturally exothermic, it is generally conducted at low temperatures to achieve the desired aroma content, resulting in a highly energy-intensive process. It is indeed necessary to minimize the loss of volatile molecules through evaporation in order to obtain a final wine with high aroma concentrations. 

Moreover, wine quality is a highly multidimensional criterion. Consequently, reducing the energy footprint without altering wine quality is a very complex problem and requires the development of breakthrough strategies to find a suitable compromise between these two key factors.

 This is the challenge that the DigitWine project aims to address.

Objectives

The overall aim of the project is to demonstrate how state-of-the-art IT tools in advanced interaction with a human decision-maker can perform the complex task of improving the environmental impact of the wine fermentation, while achieving a specified user-perceived product quality.

The two main objectives of the project and the intermediate steps to reach them, are as follows:

•  Objective 1: develop innovative fermentation management strategies to find the best compromise between energy consumption and flavour production. Based on dynamic modelling and human expertise, the control law will consist in a real-time adjustment of fermentation temperature and nitrogen addition to reach predefine targets corresponding to the multidimensional aroma profile and minimise the energy consumption.
• Objective 2: develop a functional digital twin involving a human decision maker. Whereas traditional multiobjective optimisation focuses on finding the best solution offline based on a priori defined criteria, our concept of human-centred optimisation can take into account the possible conflicting and time-varying priorities and preferences of a human decision-maker while the process is running.

^{Global structure of the DigitWine project. The arrows denote information flow between the physical process, the digital twin and the human operator.}

Expected results

A variety of outcomes are foreseen:

• Data. Several data sets will be produced: (i) constant rate fermentation experiments with liquid and gas concentration measurements of 15 previously unexplored aroma compounds, (ii) experiments with dynamic temperature changes and nitrogen additions to assess their effect on studied aromas and energy and (iii) closed-loop experiments with the digital twin in action to test and validate various control scenarios.
• Knowledge. The acquired quantitative knowledge concerning the synthesis and volatility of newly explored aroma compounds will be encapsulated in a mechanistic dynamic model, taking into account the yeast metabolic pathways and thermodynamic properties of these compounds.
• Digital twin. The developed digital twin consists of several software and hardware components. Software components will be designed to be generic and reusable as much as possible. The most innovative component will be the visualisation and human decision interface, designed to handle real-time exploration of control scenarios and multidimensional data.
• Serious game. The digital twin will be first tested off-line in an interactive “serious game” where researchers and executives from the winemaking sector will be able to explore in simulation realistic compromises between quality and energetic requirements. Beyond the scope of the project, this tool can be used to increase awareness about IT benefits in bioprocess and training of both staff and students.
• Proof of concept. The project will pave the way towards collaborations with private companies interested in implementation of advanced control tools. Beyond the winemaking sector, the developed digital twin may serve as a “success story” and a motivating example of the benefits of advanced IT in the bioprocess and food industry at large. Indeed, examples of successful digital twin implementations in these sectors remain to this day extremely scarce.

Beyond the winemaking sector, the developed digital twin can serve as a “success story” and a motivating example to illustrate the benefits of using information technology in the bioprocessing industry and the food industry more broadly.

Partners

Five research units are involved in this project:

• SayFood* (UMR INRAE, AgroParisTech, Université Paris-Saclay), Paris-Saclay Food and Bioproduct Engineering
• MIA Paris-Saclay (UMR INRAE, AgroParisTech, Université Paris-Saclay), Mathématiques et Informatique Appliquées Paris Saclay
• MISTEA (UMR INRAE, Institut Agro de Montpellier, université de Montpelier), Mathématiques, Informatique et STatistique pour l'Environnement et l'Agronomie
• Pech Rouge (UE INRAE), Unité expérimentale de Pech Rouge
• SPO (UMR INRAE, Institut Agro de Montpellier, Université de Montpellier), Sciences pour l'Œnologie 

* Coordinating Joint Research Unit

PhD students

• Omar Naboulsi

Conception d’un observateur (estimateur d’état) pour un processus de fermentation œnologique

 Publications

• From Concept to Perspective: Digital Twins of Microbial Systems, Pré-Publication – Document de Travail (2026)