The Application of LSTM Model to the Prediction of Abnormal Condition in Nuclear Power Plants

Abstract

The Long Short-Term Memory (LSTM) model is investigated in this work, as a proposed prediction method for the abnormal condition in Nuclear Power Plants (NPPs). Its advantage of processing long timeline data is utilized to overcome the limitation of the traditional Recurrent Neural Network (RNN). With the assistance of the Rolling Update (RU) method, the LSTM model is trained using historical NPP operation data to obtain the capability of predicting abnormal trends. Such prediction ability is validated using simulated accident data, which demonstrates its prediction accuracy with a loss value of 3.7 × 10−6. Moreover, it is verified in this work that LSTM can predict the trends of accidents that belong to same category but differ in certain parameters.