Research on Fast Prediction of Key Parameters of Containment Based on Time Series Deep Learning Model
-
摘要: 主蒸汽管道断裂（MSLB）事故威胁核电厂安全运行。本文基于时序深度学习模型预测核电厂非能动安全壳冷却系统（PCCS）在MSLB事故下关键安全参数随时间变化的瞬态响应。以瞬态安全参数为研究对象，数据通过线性归一化、特征标签分割预处理，使用短期数据集训练，采用长短时记忆网络（LSTM）和循环神经网络（RNN）建立单参量与多参量协同的时序深度学习模型；由多参量协同模型预测未经训练的长期数据集。研究表明：在同类事故、不同工况下，基于时序深度学习模型的预测具有适用性；基于训练短期数据来预测长期数据方法可行；使用LSTM的单参量模型或多参量协同模型的预测精度比RNN更高，基于LSTM深度学习模型能够有效、高精度快速预测MSLB事故下PCCS瞬态安全参数响应特性, 可为事故安全分析提供快速预测分析。
-
关键词:
- 主蒸汽管道断裂（MSLB） /
- 时序深度学习模型 /
- 安全分析
Abstract: The Main Steam Line Break (MSLB) accident threatens the safe operation of nuclear power plant. In this paper, the time-dependent transient response of key safety parameters of passive containment cooling system (PCCS) in nuclear power plant under MSLB accident is predicted based on time series deep learning model. The transient safety parameters are taken as the research objects. The data are preprocessed by linear normalization and feature label segmentation, trained by short-term data sets, and the time series deep learning model of single parameter and multi parameter coordination is established by using long short-term memory (LSTM) and recurrent neural network (RNN); long-term untrained data sets are predicted by a multi-parameter coordination model. The research shows that the prediction based on time series deep learning model is applicable under the same accident and different working conditions; it is feasible to predict long-term data based on short-term training data; the prediction accuracy of the single-parameter model or multi-parameter coordination model using LSTM is higher than that of RNN. The deep learning model based on LSTM can effectively, accurately and quickly predict the transient safety parameter response characteristics of PCCS under MSLB accidents, and can provide a fast prediction and analysis model for accident safety analysis. -
图 2 LSTM模型示意图
LSTM_1—第一LSTM循环计算层；LSTM_2—第二LSTM循环计算层；a、b、c、d—研究参量；y—输出结果（标签）
Figure 2. Schematic Diagram of LSTM model
图 3 单参量模型安全壳内温度预测结果
Figure 3. Prediction of Temperature in Containment by Single-parameter Model
图 4 多参量协同模型安全壳内压力预测结果
Figure 4. Prediction of Pressure in Containment by Multi-parameter Coordination Model
图 5 多参量协同模型安全壳外壁面液膜覆盖率预测结果
Figure 5. Prediction Results of Liquid Film Coverage on the Outer Wall of Containment by Multi-parameter Coordination Model
图 6 多参量协同模型安全壳外壁面液膜蒸发量预测结果
Figure 6. Prediction Results of Liquid Film Evaporation on the Outer Wall of Containment by Multi-parameter Coordination Model
表 1 初始输入参数
Table 1. Initial Input Parameters
参数 参数值 温度/℃ 16, 10 压力/MPa 0.1 初始液膜覆盖率 0.85 冷却水流量/(kg·s−1) 8.98 风速/(m·s−1) 1.37 
下载: 导出CSV
表 2 数据集划分（单参量模型）
Table 2. Data Set Classification(Single-parameter Model)
设计工况 试验工况 训练数据 800 组 测试数据 200 组 训练集 验证集 测试集 700组特征-标签 60组特征-标签 160组特征-标签
下载: 导出CSV
表 3 数据集划分（多参量协同模型）
Table 3. Data Set Classification (Multi-parameter Coordination Model)
设计工况 试验工况 训练数据 800 组 测试数据 10000 组 训练集 验证集 测试集 700组特征-标签 60组特征-标签 9960组特征-标签
下载: 导出CSV
表 4 LSTM模型的神经网络参数
Table 4. Neural Network Parameters of LSTM Model
参数 单参量模型 多参量协同模型 LSTM_1节点数 80 128 Dropout_1 0.2 0.2 LSTM_2节点数 100 Dropout_2 0.2 Dense节点数 1 3 Optimizer Adam Adam Loss function MSE MSE batch size 64 64 epochs 80 50 Optimizer—优化器；Adam—自适应矩估计优化算法；Loss function—损失函数； MSE—均方误差；batch size—批大小（一次喂入神经网络的数据量）；epochs—轮次（数据集的迭代次数）
下载: 导出CSV
表 5 多参量协同模型LSTM、RNN网络预测误差对比
Table 5. Comparison of Prediction Errors between LSTM and RNN Networks by Multi-parameter Coordination Model
模型名称 评价指标 P R E LSTM MSE $4.2 \times {10^{ - 5}}$ $1.47 \times {10^{ - 2}}$ $4.16 \times {10^{ - 8}}$ R2 0.99 0.72 0.98 RNN MSE $1.41 \times {10^{ - 3}}$ $2.5 \times {10^{ - 2}}$ $4.6 \times {10^{ - 9}}$ R2 0.65 0.52 0.98
下载: 导出CSV
-
[1] 国务院. 新一代人工智能发展规划[J]. 科技创新与生产力,2017(8): 2,121. [2] 杨红义,王端,王东东. 神经网络算法在我国核领域中的应用综述[J]. 核科学与技术,2020, 8(1): 19-34. [3] 王端,王威策,潘翠杰,等. 基于自适应BP神经网络的压水堆堆芯换料关键参数的预测方法[J]. 原子能科学技术,2020, 54(1): 112-118. doi: 10.7538/yzk.2019.youxian.0016 [4] 韩龙,周刚,孙旭升. 基于集成神经网络与模糊逻辑融合的稳压器泄漏监测方法[J]. 原子能科学技术,2014, 48(S1): 474-479. [5] 龚安,马光明,郭文婷,等. 基于LSTM循环神经网络的核电设备状态预测[J]. 计算机技术与发展,2019, 29(10): 41-45. doi: 10.3969/j.issn.1673-629X.2019.10.009 [6] 佘兢克,薛时雨,孙培伟,等. 基于深度学习的核电站事故预测及故障诊断方法[J]. 仪器仪表用户,2019, 26(12): 39-44. doi: 10.3969/j.issn.1671-1041.2019.12.010 [7] ZHANG J, WANG X L, ZHAO C, et al. Application of cost-sensitive LSTM in water level prediction for nuclear reactor pressurizer[J]. Nuclear Engineering and Technology, 2020, 52(7): 1429-1435. doi: 10.1016/j.net.2019.12.025 [8] WANG H, PENG M J, LIU Y K, et al. Remaining useful life prediction techniques of electric valves for nuclear power plants with convolution kernel and LSTM[J]. Science and Technology of Nuclear Installations, 2020, 2020: 8349349. [9] ZAGHLOUL M A S, HASSAN A M, CARPENTER D, et al. Optical sensor behavior prediction using LSTM neural network[C]//2019 IEEE Photonics Conference (IPC). San Antonio, TX, USA: IEEE, 2019. [10] LEE D, SEONG P H, KIM J. Autonomous operation algorithm for safety systems of nuclear power plants by using long-short term memory and function-based hierarchical framework[J]. Annals of Nuclear Energy, 2018, 119: 287-299. doi: 10.1016/j.anucene.2018.05.020 [11] SUN Q T, GUO Z P, CHEN W, et al. Global sensitivity analysis of the main steam line break accident by using sampling methods and surrogate models[J]. Annals of Nuclear Energy, 2021, 150: 107787. doi: 10.1016/j.anucene.2020.107787 [12] 赵炳南. 事故情况下非能动安全壳冷却系统降温降压特点与运行浅析[J]. 城市建设理论研究(电子版),2020(2): 5. [13] 胡健,温丽晶,石兴伟,等. 基于WGOTHIC程序的非能动安全壳冷却系统传热特性分析[J]. 核安全,2017, 16(4): 71-77. doi: 10.16432/j.cnki.1672-5360.2017.04.013 [14] 林诚格. 非能动安全先进核电厂AP1000[M]. 北京: 中国原子能出版社, 2010: 198. [15] YU Y, DI MAIO F, ZIO E, et al. An efficient method for passive safety systems reliability assessment[J]. Annals of Nuclear Energy, 2020, 141: 107347. doi: 10.1016/j.anucene.2020.107347 [16] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323(6088): 533-536. doi: 10.1038/323533a0 [17] KOLEN J F, KREMER S C. Gradient Flow in Recurrent Nets-the Difficulty of Learning Long-Term Dependencies [M]. USA: Wiley-IEEE Press, 2001: 237-243. [18] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. doi: 10.1162/neco.1997.9.8.1735 -
计量
- 文章访问数: 226
- HTML全文浏览量: 82
- PDF下载量: 47
- 被引次数: 0
