Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : INSA Strasbourg / Laboratoire ICube
Durée : 36 mois
Contact : franco.giustozzi@insa-strasbourg.fr
Date limite de publication : 2026-03-06

Contexte :
Industrial parks are major contributors to global energy consumption and CO2 emissions due to their high demand, heterogeneous energy users, and complex energy flows. Improving energy efficiency in these environments is therefore a key lever for achieving climate targets, reducing operational costs, and strengthening regional competitiveness, particularly in industrially dense regions such as the Upper Rhine area. Despite their importance, conventional energy management systems are typically designed as isolated solutions. They lack the capability to address large-scale challenges such as decentralized energy optimization, integration of renewable energy sources (e.g. photovoltaic systems, waste heat recovery), and coordinated load balancing across multiple stakeholders. While collaborative energy platforms offer significant potential, their real-world deployment is constrained by strict requirements regarding data security and privacy, scalability, and adaptability to changing industrial infrastructures.
The FLEX-E project1 addresses these challenges by introducing a collaborative energy optimization framework based on Federated Learning (FL). FL is a decentralized machine learning paradigm in which local entities—such as buildings, energy producers, or consumers—train models locally and share only abstracted model parameters rather than raw data. This approach enables cross-organizational learning while preserving data sovereignty, ensuring privacy, and supporting scalable deployment. In FLEX-E, this federated approach is combined with energy flow modeling based on digital twins and validation in real and planned industrial park testbeds of varying sizes. The project thus provides a unique foundation for advanced research into secure, data-driven, and collaborative energy management systems for industrial environments.

Sujet :
The increasing electrification of industry, coupled with the integration of renewable energy sources and flexible loads, has significantly increased the complexity of energy management in industrial parks. These environments are characterized by heterogeneous assets, distributed ownership, and strict requirements regarding data privacy and operational confidentiality. Traditional centralized energy management systems struggle to scale under these constraints and often fail to fully exploit collaborative optimization potentials.
This PhD project aims to advance the state of the art by developing an explainable and hybrid federated learning framework for energy optimization in industrial parks, building upon the FLEX-E project. The proposed approach combines data-driven federated learning with expert knowledge, including physics-based energy models, digital twins and knowledge graphs, to improve robustness, generalization, and trustworthiness of AI-based energy management systems.
[Full description in the attached file.]

Profil du candidat :
We are looking for a highly motivated PhD candidate with a Master (or engineer) degree (Bac+5 level) with a strong background in computer science or data science or energy systems, or a closely related field.

Formation et compétences requises :
Experience with Python and common ML frameworks (e.g. PyTorch, TensorFlow) is expected. A background or demonstrated interest in energy systems, smart grids, or industrial energy management is highly desirable. Familiarity with physical modeling, optimization, or digital twins is an advantage. Interest in explainable AI, hybrid modeling, or knowledge graphs is a plus.

Adresse d’emploi :
INSA Strasbourg.
24 Bd de la Victoire, 67000 Strasbourg.

Document attaché : 202602171120_Thesis_proposal_FLEX_E.pdf