Supervised machine learning with regression for the IRT-T reactor cooling system

Abstract

The purpose of this study is to create a machine learning model for the IRT-T reactor cooling system, which can estimate and predict the temperature difference in the secondary circuit. To do this, data was downloaded from the SCADA system, then an application was developed for converting and preprocessing this data. Then regression and classification models were constructed that evaluated the efficiency of the cooling system and its ability to predict changes in the temperature drop on heat exchangers. The main technical characteristics of the IRT-T reactor include a thermal power of 6 MW, the use of UO2 nuclear fuel in an aluminum matrix with an enrichment of 90.1%, a coolant in the form of desalinated water, tubular square-section fuel rods with external cooling, the fuel element shell material is SAV-1 alloy, and the use of five 5 IRT-1000 type heat exchangers with a total area of heat exchange of 1000 m2.