Small modular reactor reinforcement learning framework: Automating reactor core startup

Abstract

A small modular reactor (SMR) has been considered a potential alternative for achieving carbon neutrality, and therefore, an increasing number of countries are performing extensive research and development. However, this is still in the development stage, and there are several technological or economical challenges that need to be overcome. Minimizing manual operations may be considered a wise approach to reduce the number of operators. Reactor core startup, which is a manual operation, is considered as an example. A method to automate the reactor core startup via the reinforcement learning (RL) algorithm is proposed in this paper. Further, an efficient SMR dynamic simulation model that performs simulations considering the action of the RL agent to achieve states and reward is developed. The suggested SMR dynamic simulation model is validated by the data available in the existing literature. The proposed method can perform automatic reactor core startup. The proposed framework that incorporates the SMR simulator to the RL algorithm is expected to be applied to various cases for reducing manual operations and contributing to realizing a higher level of SMR automation.