An intelligent control rod movement strategy for boron-free reactor core using multi-layer perceptron machine learning model

The boron-free operation of Small Modular Reactor (SMR) core requires efficient approaches to manage excess reactivity throughout prolonged operational cycles. Adjusting Control Rods (CRs) is the only way to compensate reactivity and regulate the reactor power during the operational cycle of boron-free cores. Therefore, developing a proper CR movement strategy throughout the cycle length is crucial for boron-free cores. This study aims to apply data mining methods within machine learning approach to forecast critical CR positions at each burnup level of the boron-free core using a Multi-Layer Perceptron (MLP) model. To achieve this goal, the design of the boron-free core Control Banks (CBs) and their computations, are investigated. Furthermore, the effects of CR movement on core neutron physics parameters are considered. The regression values for training, testing, and all datasets are calculated. The results indicate that the prediction of critical CR movement strategy is properly done by the developed MLP model. The trained MLP model operates extremely quickly (less than 1 sec) and can serve as a quick support model for forecasting CR movement strategy. The forecasting results of the developed model, based on known and unknown data, verify a high correlation between forecasted and real values, demonstrating that the performance of the developed model is good and has high accuracy.