Author:
Li Xichen,Chen Xiang,She Jingke,Zhang Yifan,Wang Taizhe
Abstract
A novel deep learning model zLSTM, which evolves from Long-Short Term Memory (LSTM) with enhanced long-term processing capability, is applied to the prediction of Loss of Coolant Accident (LOCA). During the prediction process, six-dimensional multivariate coupling is established among six major system parameters after connecting each timestep with the time dimension. The demonstration experiments show that the proposed method can increase the prediction accuracy by 35.84% comparing to the traditional LSTM baseline. Furthermore, zLSTM model follows the parameter progress well at the starting stage of LOCA, which reduces the prediction error at both the beginning and the far end.
Funder
Ministry of Science and Technology of the People’s Republic of China
Reference14 articles.
1. Numerical study on influence of blowdown parameter on coolant blowdown characteristic in LOCA;Bingzheng;Atomic Energy Sci. Technol.,2020
2. GRU-CNN-Based prediction of LOCA accident condition in nuclear power plants;Fukun,2022
3. Multivariate time series prediction for loss of coolant accidents with a zigmoid-based LSTM;Gong;Front. Energy Res.,2022
4. Exploiting multi-CNN features in CNN-RNN based dimensional emotion recognition on the OMG in-the-Wild dataset;Kollias;IEEE Trans. Affect. Comput.,2020
5. Prediction of nuclear reactor vessel water level using deep neural networks;Koo,2018