Accurate prediction of key parameters of lead-bismuth reactor under accident conditions is an important content of reactor safety analysis, which is of great significance to improve the safety of the reactor under accident conditions. In this work, an optimization algorithm is used to improve the prediction performance of the Long Short Term Memory (LSTM) neural network by hyperparameter optimization, and a parameter prediction method based on the coupled optimization algorithm of multivariate LSTM neural network is proposed. For the parameter prediction problem of lead-bismuth fast reactor MARS-3 under unprotected loss of flow accident conditions, a comprehensive evaluation of the proposed method is performed using Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method after data samples generated by the sub-channel code SUBCHANFLOW. The results show that the prediction performance of the multivariate LSTM neural network coupled with the Particle Swarm optimization method is optimal, and its computational efficiency can be improved to 438 times that of SUBCHANFLOW. The relevant research results can help improve the efficiency of predicting key thermal parameters of lead-bismuth reactors and improve the emergency response capability of lead-bismuth reactors. © 2023 Yuan Zi Neng Chuban She. All rights reserved.
