Nuclear power plant core optimization power control method based on imitation learning

Abstract

The application discloses a kind of nuclear power plant core optimization power control methods based on imitation learning, based on the method of imitation learning, the core optimization power control model comprising a plurality of intelligent agents and an average layer is constructed;The output result of trained intelligent agent is averaged as the final control result.The application based on the method of imitation learning develops nuclear power plant core optimization power control, supervises and guides the training of intelligent agent to learn the decision logic of nuclear power plant power control system and operator, effectively utilizes the real data of power plant while avoiding the serious consequences caused by immature intelligent agent directly interacting with power plant.Integrated learning method is used to integrate multiple intelligent agents, so that the model can be applied to multiple nuclear power plant peak shaving, normal start-stop reactor, emergency shutdown multiple working conditions.Each intelligent agent is optimized based on Bayesian method during training, to find the best hyperparameters, improve the efficiency of network training and fault diagnosis accuracy.

Images