Development and analysis of a K-nearest-neighbor-based transient identification model for molten salt reactor systems

Tianze ZHOU; Kaicheng YU; Maosong CHENG; Zhimin DAI

doi:10.11889/j.0253-3219.2023.hjs.46.110604

Journals >NUCLEAR TECHNIQUES >Volume 46 >Issue 11 >Page 110604 > Article

NUCLEAR TECHNIQUES
Vol. 46, Issue 11, 110604 (2023)

Development and analysis of a K-nearest-neighbor-based transient identification model for molten salt reactor systems

Tianze ZHOU^1、2, Kaicheng YU^2、3, Maosong CHENG^2、3、*, and Zhimin DAI^{1、2、3、**}

Author Affiliations

¹ShanghaiTech University, Shanghai 201210, China

²Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai 201800, China

³University of Chinese Academy of Sciences, Beijing 100049, China

show less

DOI: 10.11889/j.0253-3219.2023.hjs.46.110604 Cite this Article

Tianze ZHOU, Kaicheng YU, Maosong CHENG, Zhimin DAI. Development and analysis of a K-nearest-neighbor-based transient identification model for molten salt reactor systems[J]. NUCLEAR TECHNIQUES, 2023, 46(11): 110604 Copy Citation Text

show less

Fig. 1. Development process of the system transient identification model

Download full size | View in the Article

Fig. 2. KNN-based system transient identification model for the MSR system

Download full size | View in the Article

Fig. 3. Framework of the KNN-based system transient identification model

Download full size | View in the Article

Fig. 4. RELAP5-TMSR nodalization of the MSRE

Download full size | View in the Article

Fig. 5. Hyper-parameter optimization results of the KNN-based system transient identification model

Download full size | View in the Article

Fig. 6. Confusion matrix for the KNN-based system transient identification model

Download full size | View in the Article

Fig. 7. Robustness test results

Download full size | View in the Article

Fig. 8. Hyper-parameter optimization results of the system transient identification model trained on noisy data

Download full size | View in the Article

Fig. 9. Confusion matrix for the KNN-based system transient identification model trained using noisy data on noiseless test datasets

Download full size | View in the Article

Fig. 10. Confusion matrix for the KNN-based system transient identification model trained using noisy data on noisy test datasets

Download full size | View in the Article

Fig. 11. Robustness test results of the system transient identification model trained on noisy data

Download full size | View in the Article

超参数Hyper-parameter	选取范围Range selected
邻近点数Number of neighbors	[1~10]
邻近点查找方法Algorithm to search neighbors	[“auto”, “ball_tree”, “kd_tree”, “brute”]
距离计算方法Method to calculate distance	[1, 2]
邻近点权重方式Method to weight neighbors	[“uniform”, “distance”]

Table 1. Hyper-parameters of KNN-based system transient identification model

运行工况Operation condition	详细描述Description
NOR	稳态运行Steady state
LOF_1	燃料泵故障Fuel circulating pump failure
LOF_2	冷却剂泵故障Coolant circulating pump failure
SBO	全场断电Station black out
URW_1	1根控制棒误提升Uncontrolled withdrawal of 1 control rod
URW_2	2根控制棒误提升Uncontrolled withdrawal of 2 control rods
URW_3	3根控制棒误提升Uncontrolled withdrawal of 3 control rods
FSL_hot	热段燃料盐泄漏Fuel salt leakage in hot leg
FSL_cold	冷段燃料盐泄漏Fuel salt leakage in cold leg
CSL_hot	热段冷却盐泄漏Coolant salt leakage in hot leg
CSL_cold	冷段冷却盐泄漏Coolant salt leakage in cold leg

Table 2. MSRE operation condition type

特征参数Feature parameter	单位Unit
功率Power	MW
堆芯出口温度Core outlet temperature	K
堆芯进口温度Core inlet temperature	K
二回路热段温度Secondary hot leg temperature	K
二回路冷段温度Secondary cold leg temperature	K
堆芯出口压力Core outlet pressure	kPa
堆芯进口压力Core inlet pressure	kPa
二回路热段压力Secondary hot leg pressure	kPa
二回路冷段压力Secondary cold leg pressure	kPa
燃料盐质量流量Fuel salt mass flow rate	kg·s^-1
冷却盐质量流量Coolant salt mass flow rate	kg·s^-1

Table 3. Feature parameters of identification in MSR system

真实标签

True label

预测标签Predicted label

Table 4. Binary classification results

超参数Hyper-parameter	优化结果Optimization results
邻近点数Number of neighbors	2
邻近点查找方法Algorithm to search neighbors	“auto”
距离计算方法Method to calculate distance	1
邻近点权重方式Method to weight neighbors	“distance”

Table 5. Hyper-parameters optimization results of KNN-based system transient identification model

准确率

Accuracy

精确率

Precision

召回率

Recall

F1分数

F1-score

测试结果Test results

99.99%

Table 6. Test results of KNN-based identification models

运行工况 Operation condition	F1分数 F1-score / %
NOR	100.00
LOF	100.00
LOF_2	100.00
SBO	100.00
URW_1	100.00
URW_2	99.94
URW_3	99.94
FSL_hot	100.00
FSL_cold	100.00
CSL_hot	100.00
CSL_cold	100.00

Table 7. F1-score of individual transient

准确率

Accuracy

精确率

Precision

召回率

Recall

F1分数

F1-score

测试结果Test results

97.41%

94.47%

94.61%

94.31%

Table 8. Results of system transient identification model under 30 dB SNR

超参数Hyper-parameter	优化结果Optimization results
邻近点数Number of neighbors	1
邻近点查找方法Algorithm to search neighbors	“kd tree”
距离计算方法Method to calculate distance	2
邻近点权重方式Method to weight neighbors	“uniform”

Table 9. Hyper-parameters optimization results of system transient identification model trained by data with noise

准确率

Accuracy

精确率

Precision

召回率

Recall

F1分数

F1-score

测试结果Test results

99.69%

99.26%