Coriolis Seminar - Building AI systems to secure the energy transition

18h, amphi Faurre, Ecole polytechnique

The energy transition is fundamentally a problem of scale. Replacing centralized power plants with millions of distributed assets – solar panels, wind turbines, heat pumps, batteries – changes not only how we produce power, but how we monitor, operate, and maintain it.
A single conventional power plant may rely on a small on site team and a few thousand sensors. Delivering the same power with distributed renewables requires hundreds of installations with no permanent staff on site, and hundreds of thousands of data streams. Simply replicating the operational practices of the past no longer works.
In this talk, I will show how AI systems can help secure the energy transition across the lifecycle of energy conversion systems, from manufacturing to operations and maintenance. Through real world examples from wind power, industrial refrigeration, and photovoltaics, we will explore what it takes to build AI that works in the field. We will examine the impact of rare events, the challenges of uncertainty, and how specific AI approaches are promising for renewable forecasting and power system operations.