HEAT PIPE COOLED REACTOR STATUS PREDICTION BASED ON LSTM RECURRENT NEURAL NETWORK

Abstract

In recent years, intelligent control in the field of nuclear power has been received attention. The application of deep learning methods in the field of nuclear power control is promoted. The nuclear power system is a complex structure system, the complexity of the structure lead to the more complexity of the system data. As a result, ordinary prediction methods cannot effectively reflect the relationship between time series data and equipment operating states. To solve the problem that the operating state of nuclear power equipment is difficult to accurately predict, this paper proposes a method for predicting the operating state of heat pipe cooled reactor based on long short-term memory (LSTM) neural network. Heat pipe reactor electric power, nuclear power and other parameters are predicted by LSTM, RNN, CNN. Comparing the predicted parameters obtained by the three methods, the results show that compared with RNN and CNN, the fitting performance and prediction performance of LSTM are better. The applicability of the deep learning method based on the LSTM model in the field of nuclear power plant operation safety assurance has been verified. The method based on LSTM lays the foundation for the subsequent establishment of the operator manual control intelligent auxiliary system to realize the change trend prediction of the controlled quantity.