The facts and the project
In a nuclear power station, the fuel, fabricated into rod assemblies, is loaded in a strict sequence that satisfies both economic criteria and safety requirements. Optimising the fuel loading plan is therefore a matter of considerable import. The computations for this optimisation were carried out by the CEA in 2013 using a French nuclear power station reactor. The fuel loading plan was treated in the same way as a genome: The computer analysis involved the evolution of the genome populations until those that were best suited for the two selection criteria – safety and fuel economy – were identified.
Carried out using the Corpus platform on GENCI’s Curie supercomputer, the work resulted in the development of a reference optimisation over 12 hours and using more than 1,000 processors. This represents a first step towards the creation of an operational predictive tool for power station operators.
Principal Investigator: Jean-Charles Le Pallec - CEA/DEN