Control automation in the heat-up mode of a nuclear power plant using reinforcement learning

Next-generation nuclear instrumentation and control technology is aimed at higher levels of automation and lower operation burden. In recent years, studies have been conducted to contribute to the operation of power plants using artificial intelligence technology. This paper proposes an automatic control method for plant heat-up mode using deep reinforcement-learning technology as a basic study for plant automation. First, the existing compact nuclear simulator (CNS) is expanded to enable reinforcement learning, and key elements for reinforcement learning are designed to be suitable for the heat-up mode. A deep neural-network structure and a CNS deep reinforcement-learning mechanism are then presented for automatic control. The experimental results demonstrate that deep reinforcement-learning has the potential to perform automatic control operation.