12th International Symposium on Symbiotic Nuclear Power ( ISSNP-2026)

    To address the control challenges caused by parameter coupling and response delays commonly encountered during accident transients in pressurized water reactor (PWR) nuclear power plants, this study develops a reinforcement learning (RL)-based control model using loop data from a nuclear energy system. A multi-objective control formulation incorporating temperature, pressure and flow responses is developed, and validation is conducted under a representative accident condition to analyze changes in key control parameters. The results show that the proposed model reduces the normalized deviations of temperature, pressure, and flow by 42.6%, 21.6%, and 31.2%, respectively, thereby improving the overall thermal-hydraulic response during the accident transient. This study demonstrates the feasibility of applying the proposed artificial intelligence approach to intelligent control and provides an important reference for the safe operation of nuclear energy systems.
 

Pressurized water reactor,Artificial intelligence,Reinforcement learning,Multi-objective control