In the context of an accelerated energy transition, energy systems must adapt to new constraints: widespread integration of renewable energy, growing use of electricity, decentralised production and increasingly digitalised networks. These shifts are making it essential to develop innovative solutions to guarantee the stability and efficiency of power grids. This is the context for the “AI for Smartgrids” chair at MIAI, co-run by Vincent Debusschere and Nouredine Hadjsaid, researchers at G2Elab and professors at Grenoble INP – Ense3.
Our power system is based on a systematic balance between energy production and energy consumption. Maintaining this balance is made more complex by the rise of renewable energy, intermittent by nature, and a greater use of electricity in multiple areas (mobility, heating, industry). At present, electricity represents around 20% of final energy consumption, but this could reach 50% by 2050. This transformation means that we must rethink grid management. At the same time, communication technologies and additional computing capabilities are being rolled out. These offer ways to explore new approaches using artificial intelligence.
The “AI for Smartgrids” chair is focused on three main challenges: digitalisation, decarbonisation and decentralisation. It explores the use of AI to optimise energy flows in real time, anticipate fluctuations in production and consumption, and increase network resilience. “We work with systems where decision-making is decentralised,” explains Vincent Debusschere. “This means we can make the network more resilient, by ensuring it restarts quickly in the event of outages, or by isolating certain zones so that they remain online.”
G2Elab's research uses experimental platforms to replicate different zones in the distribution grid, in which a range of scenarios can be simulated. For example, researchers study system flexibility: could an electric car not only charge itself, but also inject energy back into the grid in the event of spikes in demand? Can we optimise consumption in a district by modulating uses in line with local production? These are just a few of the questions being addressed by hybrid models combining human expertise and machine learning algorithms. “AI allows us to mitigate certain gaps when modelling energy systems,” says the researcher. “It provides predictions that can be incorrect, certainly, but it is able to learn how to manage uncertainty for them. Of course, human supervision remains key.”
Beyond the purely technological aspects, the transition to smartgrids raises economic, environmental and social questions. How can we ensure equitable access to energy while optimising available resources? What economic models should be implemented to encourage eco-friendly consumer behaviour? Experimental solutions must also be adapted to existing infrastructure, designed to last for several decades. Other parameters to consider include technology interoperability, robust monitoring systems and cybersecurity.
Beyond the debate around generative AI, which is not used in the work discussed here, G2Elab’s research as part of the MIAI chair illustrate the potential of artificial intelligence technology to respond to the energy challenges of the future. By combining modelling, optimisation and data analysis, they help shape smarter, more flexible and more sustainable grids. A key issue for the transition towards a decarbonised, resilient electricity system.
* CNRS / UGA / Grenoble INP - UGA
