Research chairs and common laboratories

SPMS laboratory and Pytheas Technology have joined forces to create a LabCom, the SPyMS, supported by the ANR, to design the electroactive materials of tomorrow. Pytheas Techonlogy is an SME based in La Ciotat, specialising in the design and production of innovative piezoelectric devices, with applications in energy harvesting, electro-hydraulics, underwater acoustics and vibration damping. SPMS laboratory has long-established expertise, particularly in electro-active materials. The LabCom is headed by Pierre-Eymeric Janolin (SPMS, Director) and Alice Aubry (Pytheas Technology, Deputy Director).

A RESEARCH INSTITUTE DEDICATED TO SMART ELECTRIC GRIDS 

The RISEGrid Institute (Research Institute for Smarter Electric Grids) was launched jointly by Supélec and EDF (Électricité de France) in December 2012 and is dedicated to the study, modelling and simulation of smart electric distribution grids and their interactions with the whole electric power system.

Research topics cover both theoretical aspects and more applicative and industrial ones. The RISEGrid Institute aims to be in perfect accordance with the strong and challenging evolutions of electric power systems all over the world: opening of electricity markets, development of decentralised production, ubiquitous information and communication means, etc.

Research studies carried out in the Institute combine the multidisciplinary aspects of smart grids (power systems, control, information systems, and telecommunication networks…). In addition, new tools for modelling and simulation of complex systems are deeply investigated.

 FOUR AXIS SCIENTIFIC PROGRAM 

Assessment of new solutions for smart electric grids

The fast and large development of decentralised electricity production has a great impact on existing distribution networks, which were designed in times when dispersed energy resources were marginal. In such a framework, the goal is to devise new solutions to increase network flexibility, not only for massive integration of decentralised production but also that of new electricity uses and applications, while still achieving high levels of reliability and security. 

Observability of the electric system

The development of new smart and automated functionalities will allow optimising the whole system, taking advantage of various flexibilities (production, consumer demand side management, storage, and electrical grid flexibility). However, for that purpose, it is required to enhance the real-time observability of the system components. The current trends rely on the use of smart and innovative signal-processing methods or data analysis algorithms. 

Information and communication systems

Information systems, communication means and infrastructures are required for the implementation of new functionalities in smart grids. The RISEGrid Institute is concerned with modelling such systems, considering their strong interactions with the electric network. Interdependencies, quality of service and network reliability are some examples of the topics encountered in this research axis, together with new approaches and tools for simulation. 

Advanced modelling and simulation

Smart grids are made of numerous interacting subsystems (electrical networks, automated meter management, centralised and decentralised production, demand side management (including smart charging for electric vehicles), storage, information systems, and telecommunication network…). All these subsystems need to be considered to obtain a fairly realistic representation of the whole system's behaviour. For that purpose, it is necessary to develop new multi-simulation tools aiming to associate dedicated subsystem simulators. New subsystems and interactions have also to be investigated, such as the automated meter management systems or the smart loading of electric vehicles.

LIAGORA, Laboratoire Intelligence Artificielle Générative pour l’Orchestration de Recherches à base d’Agents (Generative Artificial Intelligence Laboratory for Agent-based Search Orchestration)

Created by the MICS laboratory and ILLUIN TECHNOLOGY with the support of the ANR, LIAGORA aims to enable the deployment and use of Generative AI within companies while guaranteeing the use of trusted and frugal systems. Capable of creating original content, formulating answers to complex questions on a knowledge base, and acting as a versatile assistant for a wide range of professions, generative AI offers considerable potential for optimizing business processes. In particular, ILLUIN is developing a suite of products capable of processing document data (search engine for complex document corpora, document parser) or conversational data (conversational agents, conversation parser, corpus analyzer), as well as a multimodal orchestration platform that enables the design of customized AI use cases.

However, the limits of current assistants based on generative AI are still too numerous. They are too prone to hallucinations, resulting in inaccurate or incoherent responses. They use a RAG (Retrieval Augmented Generation) system to answer many queries. Yet, the RAG approach has many limitations, particularly in information retrieval for complex queries within vast specialized documentary corpora. They generally lack long-term context, leading to omissions or contradictions in extended conversations. They, too, rarely use the company's business rules, which can be formulated in various structures (documents, ontologies, databases, etc.). Last but not least, they require infrastructures that are too extensive for their large-scale use to be acceptable to most companies.

The three major challenges of this LabCom will be to develop assistants based on Generative AI:

• Efficient, by improving their relevance and diversifying possible use cases,
• Trustworthy, by reducing hallucinations through optimization of the foundation models used (LLM, VLM, etc.) or information retrieval paradigms (RAG, Agents, use of Tools, etc.),
• Frugal, to reduce the carbon, energy, and financial impact of the systems developed (through hybridization, use of small foundation models, model compression, etc.) to enable mass use and adoption with reasoned consumption.