Automating power-increase operation for small modular reactors based on task analysis with proximal policy optimization

Interest in small modular reactors (SMRs) has been growing for their enhanced safety design and operational flexibility. For their adoption, one related challenge to be resolved is the increase in the task load of operators, as SMRs are designed for multi-module operation. This challenge is further heightened during the power-increase operation, which requires continuous monitoring and manual adjustments for an extended period. To address this, this study proposes an autonomous algorithm for the tasks involved in the power-increase operation of SMRs. Utilizing an integral pressurized water reactor simulator that mimics the operation of an SMR, this study first performed a task analysis of the power-increase operation to identify candidate tasks for automation and to derive automation strategies. Based on the task analysis results, an algorithm was designed by combining a deep reinforcement learning–based system with a rule-based system, specifically proximal policy optimization and if-then rules, respectively. Experimental results showed that the autonomous algorithm successfully managed the power-increase operation, achieving 100 % reactor power while staying within operational constraints. Additionally, the developed algorithm was implemented and visualized through a graphical user interface, providing real-time monitoring of key parameters during the power-increase operation.