An accident diagnosis method of pressurized water reactor based on BI-LSTM neural network

Abstract

Due to the lack of accident data, previous researchers used accident data obtained from system code simulation to train and verify the machine learning models. In this study, a long short-term memory (LSTM) neural network was used to construct a pressurized water reactor (PWR) accident diagnosis model. The model was trained and tested using simulation data of a PWR based on the deterministic analysis method and uncertainty analysis method. The accuracy of the LSTM model based on deterministic analysis and uncertainty analysis are 96.67% and 98.77%, respectively. The LSTM model developed in the present study is superior to traditional mathematical statistics and support vector machine (SVM) models in accident diagnosis, especially for diagnoses with uncertain data. The results of this study show that the LSTM model has practical value in reactor accident diagnosis.