基于POD及神经网络的蒸汽发生器传热管内温度场的快速重构技术研究

张贺; 梁彪; 王博; 谭思超; 韩蕊; 李江宽; 田瑞峰

doi:10.13832/j.jnpe.2024.070047

基于POD及神经网络的蒸汽发生器传热管内温度场的快速重构技术研究

doi: 10.13832/j.jnpe.2024.070047

张贺^{1, 2,},
梁彪^{1, 2},
王博^{1, 2, ,},
谭思超^{1, 2},
韩蕊^{1, 2},
李江宽^{1, 2},
田瑞峰^{1, 2}

1.
哈尔滨工程大学黑龙江省核动力装置性能与设备重点实验室，哈尔滨，150001
2.
哈尔滨工程大学核科学与技术学院，哈尔滨，150001

基金项目: 中核集团领创基金项目（CNNC-LCKY-202251）

详细信息

作者简介:
张　贺（1998—），男，硕士研究生，现主要从事核动力系统智能化方面的研究，E-mail: he123@hrbeu.edu.cn

通讯作者:
王　博，E-mail: bowang@hrbeu.edu.cn

中图分类号: TL334
计量
- 文章访问数: 99
- HTML全文浏览量: 14
- PDF下载量: 1
- 被引次数: 0
出版历程
- 收稿日期: 2024-07-17
- 修回日期: 2024-07-29
- 网络出版日期: 2025-01-23
- 刊出日期: 2025-04-02

Research on Rapid Reconstruction Technology of Temperature Field in Heat Transfer Tube of Steam Generator Based on POD and Neural Network

Zhang He^{1, 2
,},
Liang Biao^{1, 2},
Wang Bo^{1, 2
, ,},
Tan Sichao^{1, 2},
Han Rui^{1, 2},
Li Jiangkuan^{1, 2},
Tian Ruifeng^{1, 2}

1.
Heilongjiang Provincial Key Laboratory of Nuclear Power System & Equipment, Harbin Engineering University, Harbin, 150001, China
2.
College of Nuclear Science and Technology, Harbin Engineering University, Harbin, 150001, China

摘要

摘要: 套管式直流蒸汽发生器的二次侧流域涉及到复杂的两相流动，数值模拟方法虽然能够精准地进行仿真计算，但其计算速度缓慢，对于多工况、瞬态条件下的计算耗时长，计算资源占用较大。模型降阶是一种将复杂系统转化为一个近似简化系统的方法，能够在保留原系统主要特征的同时实现快速计算。本研究采用本征正交分解（POD）方法对换热管内温度场进行模型降阶，截取有限模态对原复杂系统进行投影获取模态系数，应用神经网络方法捕捉长短期时序模态系数分布规律。研究结果表明，预测重构温度场误差在15%范围内，且预测速度相较于数值模拟方法能够提升4个数量级。因此，本研究建立的模型降阶耦合神经网络的预测方法能够用于套管内温度场的快速预测，为其内部热工水力分析提供支撑。
- 温度场 /
- 模型降阶 /
- 神经网络 /
- 快速预测
Abstract: The secondary flow region of a casing once-through steam generator involves complex two-phase flow. Although numerical simulation methods can achieve precise simulation calculations, they are slow and time-consuming for multi-condition and transient calculations, and consume significant computational resources. Model reduction is a method that transforms a complex system into an approximately simplified system, enabling rapid calculations while retaining the main characteristics of the original system. This study employs the Proper Orthogonal Decomposition (POD) method to reduce the model of the temperature field inside the heat exchange tubes, capturing the modal coefficients by projecting the original complex system onto a limited number of modes. A neural network method is applied to capture the distribution patterns of short- and long-term time series modal coefficients. The research results indicate that the error in predicting the reconstructed temperature field is within 15%, and the prediction speed is improved by four orders of magnitude compared to numerical simulation methods. Therefore, the prediction method established in this study, which couples model order reduction with neural networks, can be utilized for the rapid prediction of the temperature field within the casing, providing support for internal thermal-hydraulic analysis.
- Temperature field /
- Model reduction /
- Neural network /
- Rapid prediction

HTML全文

图 1 物理计算域选取

Figure 1. Selection of Physical Computing Domain

下载: 全尺寸图片幻灯片

图 2 网格划分

Figure 2. Computational Domain Meshing

下载: 全尺寸图片幻灯片

图 3 网格无关性验证

Figure 3. Mesh Independence Verification

下载: 全尺寸图片幻灯片

图 4 出口温度数值模拟及系统仿真计算结果对比验证

Figure 4. Comparison and Validation of Numerical Simulation and System Simulation Results for Outlet Temperature

下载: 全尺寸图片幻灯片

图 5 出口温度计算相对误差

Figure 5. Relative Error of Calculated Outlet Temperature

下载: 全尺寸图片幻灯片

图 6 温度场预测重构算法框架

Figure 6. Temperature Field Prediction Reconstruction Algorithm Framework

下载: 全尺寸图片幻灯片

图 7 残差块的基本结构

Figure 7. Basic Structure of Residual Block

下载: 全尺寸图片幻灯片

图 8 TCN-GRU网络架构

Figure 8. TCN-GRU Network Architecture

下载: 全尺寸图片幻灯片

图 9 数据集设置

Figure 9. Dataset Configuration

下载: 全尺寸图片幻灯片

图 10 模态系数能量分布

Figure 10. Modal Coefficient Energy Distribution

下载: 全尺寸图片幻灯片

图 11 升功率运行工况预测结果

Figure 11. Predicted Results of Power Increase Operating Conditions

下载: 全尺寸图片幻灯片

图 12 稳定运行工况预测结果

Figure 12. Predicted Results of Stable Operating Conditions

下载: 全尺寸图片幻灯片

图 13 1100 s重构结果及误差分布

Figure 13. Reconstruction Results and Error Distribution at 1100 s

下载: 全尺寸图片幻灯片

图 14 1300 s重构结果及误差分布

Figure 14. Reconstruction Results and Error Distribution at 1300 s

下载: 全尺寸图片幻灯片

图 15 1500 s重构结果及误差

Figure 15. Reconstruction Results and Error Distribution at 1500 s

下载: 全尺寸图片幻灯片

图 16 8100 s重构结果及误差

Figure 16. Reconstruction Results and Error Distribution at 8100 s

下载: 全尺寸图片幻灯片

图 17 8500 s重构结果及误差

Figure 17. Reconstruction Results and Error Distribution at 8500 s

下载: 全尺寸图片幻灯片

表 1 部分时间步POD及预测重构最大相对误差

Table 1. Maximum Relative Error in POD and Prediction Reconstruction at Certain Time Steps

时间/s	POD重构误差	预测重构误差
1100	0.034	0.033
1300	0.058	0.058
1500	0.099	0.103
8100	0.134	0.136
8500	0.138	0.134

下载: 导出CSV

表 2 实验环境配置

Table 2. Configuration of Experimental Environment

名称	配置信息
操作系统	Windows 10 22H2
开发语言	Python 3.9.17
框架	Pytorch 1.12.0 + cuda 11.6
中央处理器（CPU）	Intel(R) Xeon(R) CPU E5-2696 v4 @ 2.20 GHz*2
图形处理器（GPU）	GeForce RTX 3060 (12G)
内存	128G

下载: 导出CSV

参考文献(14)

[1]	谭思超,李桐,刘永超,等. 关于人工智能在核能领域应用的若干思考[J]. 核动力工程,2023, 44(2): 1-8.
[2]	路宽,张亦弛,靳玉林,等. 本征正交分解在数据处理中的应用及展望[J]. 动力学与控制学报,2022, 20(5): 20-33.
[3]	丁鹏,陶文铨. 一种预测流动和传热问题的快速算法[J]. 西安交通大学学报,2007, 41(3): 271-273. doi: 10.3321/j.issn:0253-987X.2007.03.002
[4]	胡和敏,卜永东,张凯,等. 直接空冷凝汽器流动和传热的降维分析[J]. 现代电力,2013, 30(2): 31-36. doi: 10.3969/j.issn.1007-2322.2013.02.007
[5]	杨涛,赵鹏程,赵亚楠,等. 基于CFD的铅铋快堆上腔室降阶热分层模型开发[J]. 核动力工程,2023, 44(2): 48-53.
[6]	WANG B, CHEN B W, WANG G Q, et al. Back propagation (BP) neural network prediction and chaotic characteristics analysis of free falling liquid film fluctuation on corrugated plate wall[J]. Annals of Nuclear Energy, 2020, 148: 107711. doi: 10.1016/j.anucene.2020.107711
[7]	杨迪,段承杰,丁鹏,等. 参数化流动传热问题的模型降阶方法研究[J]. 原子能科学技术,2024, 58(7): 1440-1451.
[8]	LIU G, HU W J, HAO S Y, et al. A fast computational method for internal temperature field in Oil-Immersed power transformers[J]. Applied Thermal Engineering, 2024, 236: 121558. doi: 10.1016/j.applthermaleng.2023.121558
[9]	PENG X J, CHEN Z D, ZHANG A M, et al. Digital twin temperature field prediction of laser powder bed fusion through proper orthogonal decomposition with radial basis function[J]. Materials Today Communications, 2024, 38: 107883. doi: 10.1016/j.mtcomm.2023.107883
[10]	XU W L, ZHONG W Q, ZHOU G W, et al. Optimization of air distribution and coal blending in pulverized coal boilers for high-temperature corrosion prevention based on POD reduced-order modeling[J]. Applied Thermal Engineering, 2024, 255: 123705. doi: 10.1016/j.applthermaleng.2024.123705
[11]	ZHAO X Y, CHEN X Q, GONG Z Q, et al. A hybrid method based on proper orthogonal decomposition and deep neural networks for flow and heat field reconstruction[J]. Expert Systems with Applications, 2024, 247: 123137. doi: 10.1016/j.eswa.2024.123137
[12]	ZHU W X, WU Y L, CAO Z F, et al. Real-time reconstruction of 3D transient non-uniform temperature field for thermal protection system based on machine learning[J]. Aerospace Science and Technology, 2024, 151: 109241. doi: 10.1016/j.ast.2024.109241
[13]	闵光云,姜乃斌. POD和DMD方法对燃料棒流致振动特性的分析[J]. 原子能科学技术,2024, 58(10): 2162-2172.
[14]	LEA C, FLYNN M D, VIDAL R, et al. Temporal convolutional networks for action segmentation and detection[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu: IEEE, 2017: 1003-1012.