A Data-driven Incident Detection Method for the Safe Operation of Molten Salt Fast Reactors

This paper presents an innovative incident detection method aiming at improving the safety and reliability of the Molten Salt Fast Reactor power plant, focusing on operational scenarios involving some deviations from normal operational conditions. The first part of the paper is devoted to presenting and discussing a data-driven incident detection and classification methodology (based on the kNN algorithm), which aims at identifying abnormal plant conditions thanks to a continuous monitoring of some measurable system parameters and variables (e.g., the molten salt temperatures in the secondary circuit). Then, the incident detection algorithm proposed is trained with a set of simulated scenarios featured by deviations of the main plant parameters from their nominal values. The data-driven model is then assessed considering increasingly complex incident classification rules, showing good performances of the model in detecting plant anomalies (with a classification accuracy ranging between 89% and 99%). Finally, for a certain abnormal state of the system, a fault detection method is sketched to estimate the probability that certain combinations of physical parameters could lead the system in that abnormal state.