The application of time series deep learning model to the fast prediction of parameters in the MSLB accident

Abstract

The Main Steam Line Break Accident (MSLB) threatens the safe operation of nuclear power plants. The transient safety parameters of the Passive Containment Cooling System (PCCS) in the MSLB accident are predicted by the time series deep learning model based on LSTM or RNN. In the dataset preprocessing, the linear normalization and two of feature label segmentation methods are used. The neural network models and the other models, namely the GBRT, RF and SVR, are used to construct the multi-parameter time series deep learning models to predict the transient safety parameters. The performance of the different models and the impact of two segmentation methods on models are compared, and it is obtained that the prediction accuracy of LSTM is higher than that of other models.