딥러닝을 이용한 비정상 상태시 원자로 트립변수 예측

Abstract

When an abnormal state occurs in a nuclear power plant, the operator must be aware of the situation based on a lot of information, which can increase human error and miss the timing of the action. These can eventually cause a reactor trip. Therefore, in order to prevent unexpected reactor trip, the variables that cause the reactor trip were predicted using artificial intelligence. The variables that cause the reactor trip in a nuclear power plant include those related to pressurizer, steam generator, neutron flux, and so on. Among these variables, the pressurizer water level and pressure were predicted in this thesis. Applied artificial intelligence methods to predict the trip variables are Long Short-Term Memory (LSTM) with attention mechanism. The performance of the developed prediction model was compared with the basic LSTM to which the attention mechanism was not applied. If information on the future behavior of the trip variables is provided to the operator through the developed prediction model, it is expected that it will be possible to reduce the human error of the operator and prevent the reactor trip in abnormal situations.