Don’t miss ANITI’s upcoming scientific seminar, on 22nd October at 3.pm, about, in IRIT.
Machine learning for critical systems management: promises and challenges for large power systems.
Patrick Panciatici, from RTE will be presenting.
Abstract
Large power systems are the most complex machines ever built by man and the energy transition is transforming them in truly cyber-physical systems of systems. Managing these systems requires solving very complex problems: detection, control and optimization. Most of them are NP-Hard problems and ad-hoc heuristics taking into account some physical characteristics, are used since the very beginning to find implementable solutions based on temporal and spatial separations. Large power systems are critical: large-scale and long-lasting power outages (blackout) have a considerable impact on our modern societies.
We must have certificates in our decision-making processes. Decisions must ensure that the system remains in a safe state even in the case of credible contingencies.
Machine learning proposes promising approaches to solve very complex decision making problems, but managing a power system is more complex than playing Go.
Finding certificates, bounds, confidence intervals remains essential even if we want to use ML. We will present how we propose to combine ML with more classical optimization/control approaches to speed up computation for online decision making or for planning future systems.
About the speaker
Patrick Panciatici graduated from Supelec, joined EDF R&D in 1985 then he joined RTE (French Transmission System Operator) in 2003 and participated in the creation of the internal R&D department of RTE. He has more than 30 years’ experience in the field of R&D for transmission systems. Presently, as a senior scientific advisor, he inspires, coordinates and supervises long term research activities in RTE. He is a member of CIGRE, a Fellow of IEEE and an Emerite member of SEE.
English 
French