GRU-CNN-Based Prediction of LOCA Accident Condition in Nuclear Power Plants
Fukun Chen, Xiaomeng Dong, Luo Yicheng · 2022
In view of the development of Deep Learning and the emergence of the demand for intelligent reactor control, the GRU (Gated Recurrent Unit), a kind of cyclist neural network, is used in this paper to predict the working condition parameters of LOCA (Loss of Coolant Accident). The experiments show, the excellent ability of GRU to capture the information of long time series is suitable for the prediction of accident condition in nuclear power plants. Besides, for accident predictions with few data set, two methods of accident trend prediction are proposed in this paper to better show the short-term development trend of accident condition. In addition, considering the feature extraction ability of CNN (Convolutional Neural Networks), we try to fuse CNN and GRU to predict parameters under different break sizes, the results show that it improves the generalization of the model.