The MathTech challenge is based on the tried-and-tested SEME (mathematics/enterprise study week) recipe proposed by AMIES, with the addition of debriefing sessions prepared in conjunction with the training organization iWips, which supports EDMH in the supervision of the entire MathTech cross-disciplinary training program.
This edition of the MathTech challenge was designed and supervised by a group of researchers from Groupe Atlantic and EDF lab. They worked hard in advance to formalize ambitious and stimulating topics that were nonetheless accessible to exploration over the space of a week. It's worth noting that the two EDF lab researchers involved are also professors in Orsay's mathematics department. Their good knowledge of the characteristics of the mathematics student population in Paris-Saclay, combined with the remarkable collective involvement of Groupe Atlantic researchers throughout the week of the challenge, greatly contributed to the success of this 2023 edition.
The topics
Two topics at the heart of Groupe-Atlantic's business issues were proposed to the EDMH doctoral students tackling this 2023 challenge. The first topic also made sense for EDF lab. The doctoral students were able to draw on data provided by both companies to propose a draft solution to the following question
1. How can we reduce a home's electricity bill without compromising the comfort of its occupants?
The second topic was more of a mathematical modeling exercise, and did not require any data processing. The underlying practical question was
2. Is it possible to use a fleet of electric water heaters for energy storage while guaranteeing domestic hot water comfort?
The jury
The jury was made up of academic representatives from EDMH, expert researchers from the two companies who proposed the above topics, and iWips trainers. In the unanimous opinion of the jury members, the presentations made by the doctoral student teams were of the highest quality, and at times even impressive.
The solutions
Here is an overview of the solutions put forward by the teams who were judged winners by the jury.
1. For the first topic, the winning team put forward a method based on reinforcement learning. The idea is to use meteorological databases coupled with a simulator to assess the impact of proposed adjustments.
2. For the second topic, the winning solution was ingenious. The modeling of the thermal dynamics of the water-heater park is simplified by a probabilistic mean-field approach, enabling us to discuss the existence of optimal optimization strategies.
Video
MathTech challenge 2023 - Testimonials
