NUCLEAR TECHNIQUES, Volume. 47, Issue 1, 010603(2024)
Method of predicting transient thermal hydraulic parameters of the core based on the gated recurrent unit model of soft attention
Figures & Tables(11)
Fig. 4. GRU prediction topological structure based on the attention mechanism
Fig. 5. Comparison of time series before and after denoising (a) Un-denoised and denoised mass flow time series, (b) Un-denoised and denoised temperature time series
Fig. 6. Relationship between the super parameter of step length and average relative error (a) Mass flow time series, (b) Temperature time series
Fig. 7. Prediction results (a) Mass flow time series, (b) Temperature time series
Table 1. Comparison of neural network prediction results
View tableView in ArticleTable 1. Comparison of neural network prediction results
预测变量
Predictive variables 神经网络算法
Neural network algorithm
平均相对误差
Mean relative error
质量流量
Mass flow
自适应BP神经网络 Adaptive BP neural network 0.062 9（连续预测 Continuous prediction） 自适应RBF神经网络 Adaptive RBF neural network 0.059 2（连续预测 Continuous prediction） 注意力+GRU Attention+GRU 0.042 9 温度
Temperature
自适应BP神经网络 Adaptive BP neural network 0.011 3（连续预测Continuous prediction） 自适应RBF神经网络 Adaptive RBF neural network 0.008 8（连续预测 Continuous prediction） 注意力+GRU Attention+GRU 0.004 2
Table 2. Comparison of DB8 wavelet denoising ablation experiments
View tableView in ArticleTable 2. Comparison of DB8 wavelet denoising ablation experiments
预测变量Predictive variables 质量流量Mass flow 温度Temperature 原始数据Raw data 0.051 6 0.004 7 去噪数据Denoising data 0.042 9 0.004 2
Table 3. Comparison of attention mechanism ablation experiments
View tableView in ArticleTable 3. Comparison of attention mechanism ablation experiments
预测变量
Predictive variables
质量流量
Mass flow
温度
Temperature
GRU基准网络 GRU benchmark network 0.049 0 0.004 3 GRU基准网络+Attention GRU benchmark network + Attention 0.042 9 0.004 2
Tools
Get Citation
Copy Citation Text
Siqi CHUN, Huan FENG, Anni ZHANG, Pengcheng ZHAO. Method of predicting transient thermal hydraulic parameters of the core based on the gated recurrent unit model of soft attention[J]. NUCLEAR TECHNIQUES, 2024, 47(1): 010603
Paper Information
Category: Research Articles
Received: May. 29, 2023
Accepted: --
Published Online: Mar. 7, 2024
The Author Email: ZHAO Pengcheng (赵鹏程)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20