Thesis defense - A distributed approach based on a multi-agent system for energy optimization of a DC distribution microgrid

We are delighted to invite you to Mr. Karem Hafsi's thesis defense, which will take place on Tuesday, January 30, 2024 at 1:30 pm in Amphitheater A042 of Esisar (50 Rue Barthélémy de Laffemas, AB Building).

This thesis in entitled: "A distributed approach based on a multi-agent system for energy optimization of a DC distribution microgrid"

This thesis was conducted within the Laboratory of Conception and Integration of Systems as part of a CIFRE thesis engaged with UBIANT company (November 2022) and which will conclude under a contract with Schneider Electric.
Karem's thesis was co-directed by Denis Genon-Catalot and Jean-Marc Thiriet from Gipsa-Lab.

The jury members will be:
Mme Anne-Lise GEHIN, University of Lille, Rapporteuse (Reporter)
M. Thierry VAL, University of Toulouse Jean-Jaurès, Rapporteur
M. Christophe BATARD, University of Nantes, Examinateur (Examiner)
M. Bertrand RAISON, University Grenoble Alpes, Examinateur
M. Stéphane FOLLIC, Schneider Electric, Invité (Guest)
M. Jean-Marc THIRIET, University Grenoble Alpes, Directeur de thèse (Thesis Director)
M. Denis GENON-CATALOT, University Grenoble Alpes, Co-directeur de thèse (Co-Thesis Director)

Abstract:

The scientific contribution of this thesis focuses on addressing specific challenges related to the stability and optimization of direct current (DC) distribution networks. This research aims to develop control strategies and advanced optimization algorithms to maximize the integration of renewable energy sources in DC while minimizing energy losses in order to achieve self-sufficiency. In pursuit of this objective, we have proposed in this work a hardware and a software architecture. The hardware architecture relies on two DC electrical distribution technologies, namely Power over Ethernet (PoE) and Power over Data Lines (PoDL). With these two technologies, it is possible to achieve at least a 7% improvement in energy efficiency when the energy source is in DC, such as photovoltaic energy or energy stored in batteries. Integrating these technologies into buildings creates a more flexible distribution system, allowing for the incorporation of loads in microgrid stabilization. PoE powers devices via Ethernet, while PoDL enables IP communication and power injection through the existing electrical wiring infrastructure. The software architecture that we have proposed in this work is based on reactive communication among the actors of the DC network, thus ensuring stability and safety. The method relies on a coordinated multi-agent system for context-aware decisions, maximizing the benefits of the DC network while ensuring stable and energy-efficient operation.

Implementing a reactive electrical distribution system is a major challenge. The decentralized nature of this system requires a communication protocol capable of meeting latency and flexibility requirements. Therefore, after evaluating several protocols, we have validated that the Data Distribution Service (DDS) protocol stands out for its real-time performance, offering well-bounded and controlled latency. Due to its distributed structure and advanced data management capabilities, DDS can ensure dependable and predictable real-time communication. In addition to the reactive communication protocol, we have used a multi-agent approach for its benefits like uncertainty management, flexibility, scalability, and distributed decision-making. However, their implementation in a context where microgrid reactivity and stability are essential represents a particular challenge. Within this specific context, we have developed a coordination-based multi-agent system architecture with well-controlled interactions and latencies. Furthermore, we have proposed an optimization method based on the Hamiltonian cycle from graph theory to optimize the latencies of the agents in their tasks. The proposed multi-agent approach and optimization algorithms aim to simultaneously address the reactivity requirements of the control system and the challenges related to optimizing energy management in the DC microgrid.

Keywords: Distributed system, DC microgrid, Multi-agent systems, Renewable energy integration, Optimization.

If you can't attend the defense, you will be able to follow Karem presentation through the following Zoom link:
https://univ-grenoble-alpes-fr.zoom.us/j/99266333052?pwd=bmRUUEdJbXVLZV…
secret code : 521522

Looking forward to see you for this important event in Karem career.