External Temperature Effects on Thermal Features of Reactor Coolant System and Power Level Estimation~1

摘要

We presented analytics of infrared sensor measurements of the secondary cooling system of HFIR spanning multiple cycles, which showed a strong dependence of the derived features of temperature difference and aggregate fan activity on the external temperature, as indicated by linear and non-linear regressions. Consequently, ML methods used for HFIR power level estimation that are trained with a single cycle data have large estimation errors when applied to other cycles, thereby indicating their certain limitations. When training data from multiple cycles is available, we illustrated that this relationship can be exploited in principle to customize ML methods to achieve significantly reduce error. These initial results indicate the potential benefits of incorporating the external temperature information into ML methods for power level estimation, and more efforts are needed to understand and fully exploit the underlying relationships.