地下高压大口径输气管道泄漏扩散时序预测模型

doi:10.16265/j.cnki.issn1003-3033.2025.12.0676

摘要/Abstract

摘要：

为解决地下高压大口径输气管道泄漏后泄漏区域空间浓度信息不明、事故未来态势不清等问题,融合机器学习降维与时序预测方法,构建高压大口径输气管道泄漏扩散预测模型。首先,基于计算流体动力学方法,构建多工况输气管道泄漏扩散浓度场数据集;其次,利用该数据集分别对预测模型中降维模块与时序预测模块进行超参数优化及训练;最后,基于独立测试集,评估高压大口径输气管道泄漏扩散预测模型预测精度,并分析不同预测时长下预测模型误差。结果表明:预测模型在测试集上平均绝对误差为0.000 5,平均绝对百分比误差为6.82%,且平均绝对百分比误差在多种预测时间步长下均低于14%。

关键词: 高压大口径输气管道, 天然气, 泄漏扩散, 时序预测, 降维模型

Abstract:

In order to address the challenges of unclear spatial concentration distribution and uncertain future evolution in high-pressure large-diameter gas pipeline leakage scenarios, a predictive model for gas leakage dispersion was proposed by integrating machine learning-based dimensionality reduction and time series forecasting methods. Firstly, a multi-condition dataset of gas leakage concentration fields was generated using computational fluid dynamics simulations. Subsequently, the dimensionality reduction module and time series forecasting module of the predictive model were separately optimized and trained using this dataset. Finally, the model's predictive accuracy was evaluated on an independent test set, and the prediction errors under various forecast horizons were analyzed. The results show that the model achieves a mean absolute error of 0.000 5 and a mean absolute percentage error (mAPE) of 6.82% on the test set, with the mAPE remaining below 14% across different prediction time steps.

Key words: high-pressure large-diameter gas pipeline, natural gas, leakage and dispersion, temporal prediction, reduction model

中图分类号:

X944.4

韩子齐, 徐童, 蔡继涛, 张思琪, 张力. 地下高压大口径输气管道泄漏扩散时序预测模型[J]. 中国安全科学学报, 2025, 35(12): 154-163.

HAN Ziqi, XU Tong, CAI Jitao, ZHANG Siqi, ZHANG Li. Time-series prediction model for leakage dispersion of buried high-pressure large-diameter gas pipelines[J]. China Safety Science Journal, 2025, 35(12): 154-163.

图/表 8

图1

表1

图2

图3

图4

图5

表2

图6

参考文献 27

[1]	彭善碧, 罗雪. 埋地掺氢天然气管道泄漏扩散数值模拟研究[J]. 中国安全科学学报, 2024, 34(3): 63-69. doi: 10.16265/j.cnki.issn1003-3033.2024.03.1904
	PENG Shanbi, LUO Xue. Numerical simulation of leakage and diffusion in buried hydrogen-blended natural gas pipeline[J]. China Safety Science Journal, 2024, 34(3): 63-69. doi: 10.16265/j.cnki.issn1003-3033.2024.03.1904
[2]	李云涛, 张振永, 刘玉卿, 等. 天然气管道全管径断裂事故影响范围研究[J]. 中国安全科学学报, 2020, 30(9): 149-154.
	LI Yuntao, ZHANG Zhenyong, LIU Yuqing, et al. Influence range of full bore rupture accidents of natural gas pipelines[J]. China Safety Science Journal, 2020, 30(9): 149-154.
[3]	YAN Yuting, DONG Xiaoqiang, LI Junming. Experimental study of methane diffusion in soil for an underground gas pipe leak[J]. Journal of Natural Gas Science and Engineering, 2015, 27: 82-89. doi: 10.1016/j.jngse.2015.08.039
[4]	BONNAUD C, CLUZEL V, CORCOLES P, et al. Experimental study and modelling of the consequences of small leaks on buried transmission gas pipeline[J]. Journal of Loss Prevention in the Process Industries, 2018, 55: 303-312. doi: 10.1016/j.jlp.2018.06.010
[5]	WANG Xuemei, TAN Yufei, ZHANG Tiantian, et al. Numerical study on the diffusion process of pinhole leakage of natural gas from underground pipelines to the soil[J]. Journal of Natural Gas Science and Engineering, 2021, 87:DOI:10.1016/j.jngse.2020.103792.
[6]	BU Fanxi, LIU Yang, LIU Yongbin, et al. Leakage diffusion characteristics and harmful boundary analysis of buried natural gas pipeline under multiple working conditions[J]. Journal of Natural Gas Science and Engineering, 2021, 94:DOI:10.1016/j.jngse.2021.104047.
[7]	王岩, 黄弘, 黄丽达, 等. 土壤大气耦合的燃气泄漏扩散数值模拟[J]. 清华大学学报:自然科学版, 2017, 57(3): 274-280.
	WANG Yan, HUANG Hong, HUANG Lida, et al. Numerical simulations of leakage gas dispersion based on soil and atmosphere coupling[J]. Journal of Tsinghua University: Science and Technology, 2017, 57(3): 274-280.
[8]	程凡, 付明, 李垣志, 等. 土壤地下空间耦合下天然气泄漏扩散数值模拟[J]. 科学技术与工程, 2023, 23 (22): 9746-9753.
	CHENG Fan, FU Ming, LI Yuanzhi, et al. Numerical simulation of natural gas leakage and diffusion in soil and underground space coupling[J]. Science Technology and Engineering, 2023, 23 (22): 9746-9753.
[9]	王向阳, 杜美萍, 汪彤, 等. 埋地燃气管道泄漏扩散过程数值模拟[J]. 中国安全科学学报, 2018, 28 (2): 45-50.
	WANG Xiangyang, DU Meiping, WANG Tong et al. Numerical simulation of leakage of gas from buried pipeline and its diffusion process[J]. China Safety Science Journal, 2018, 28 (02): 45-50.
[10]	何沫, 唐雨, 李懿, 等. 含硫输气管道泄漏扩散多气团叠加计算模型[J]. 中国安全科学学报, 2016, 26(5): 118-123.
	HE Mo, TANG Yu, LI Yi, et al. Multiple-puff superposition calculation model for release and dispersion of sour gas form pipeline[J]. China Safety Science Journal, 2016, 26(5): 118-123.
[11]	WANG Binbin, CHEN Liqiong, LIN Wanxin, et al. Research on gas diffusion of natural gas leakage based on Gaussian plume model[J]. Arabian Journal of Geosciences, 2022, 15:DOI:10.1007/s12517-022-09922-6.
[12]	PADULA G, GIRFOGLIO M, ROZZA G. A brief review of reduced order models using intrusive and non-intrusive techniques[J]. Proceedings in Applied Mathematics and Mechanics, 2024, 24:DOI:10.1002/pamm.202400210.
[13]	HASEGAWA K, FUKAMI K, MURATA T, et al. Data-driven reduced order modeling of flows around two-dimensional bluff bodies of various shapes[C]. Fluids Engineering Division Summer Meeting, 2019:DOI:10.1016/j.ijheatfluidflow.2021.108816.
[14]	GEELEN R, WRIGHT S, WILLCOX K. Operator inference for non-intrusive model reduction with quadratic manifolds[J]. Computer Methods in Applied Mechanics and Engineering, 2023, 403:DOI:10.1016/j.cma.2022.11571.
[15]	EIVAZI H, GUASTONI L, SCHLATTER P, et al. Recurrent neural networks and Koopman-based frameworks for temporal predictions in a low-order model of turbulence[J]. International Journal of Heat and Fluid Flow, 2021, 90:DOI:10.1016/j.ijheatfluidflow.2021.108816.
[16]	HE Xu, KONG Depeng, YU Xirui, et al. Prediction model for the evolution of hydrogen concentration under leakage in hydrogen refueling station using deep neural networks[J]. International Journal of Hydrogen Energy, 2024, 51: 702-712. doi: 10.1016/j.ijhydene.2022.12.102
[17]	HE Xu, KONG Depeng, YANG Guodong, et al. Hybrid neural network-based surrogate model for fast prediction of hydrogen leak consequences in hydrogen refueling station[J]. International Journal of Hydrogen Energy, 2024, 59: 187-198. doi: 10.1016/j.ijhydene.2024.01.328
[18]	RAJ N A, TAFTI D, MURALIDHAR N. Comparison of reduced order models based on dynamic mode decomposition and deep learning for predicting chaotic flow in a random arrangement of cylinders[J]. Physics of Fluids, 2023, 35:DOI:10.1063/5.0153186.
[19]	SOLERA-RICO A, SANMIGUEL C, GÓMEZ-LÓPEZ M, et al. β -Variational autoencoders and transformers for reduced-order modelling of fluid flows[J]. Nature Communications, 2024, 15:DOI:10.1038/s41467-024-45578-4.
[20]	GB 50494—2009, 城镇燃气技术规范[S].
	GB 50494-2009, Technical code for city gas[S].
[21]	PETERSON E W, HENNESSEY J P. On the use of power laws for estimates of wind power potential[J]. Journal of Applied Meteorology and Climatology, 1978, 17(3): 390-394.
[22]	SHIH T H, LIOU W W, SHABBIR A, et al. A new k-ε eddy viscosity model for high reynolds number turbulent flows[J]. Computers & fluids, 1995, 24(3): 227-238. doi: 10.1016/0045-7930(94)00032-T
[23]	OLIVEIRA P J, ISSA R I. An improved PISO algorithm for the computation of buoyancy-driven flows[J]. Numerical Heat Transfer, Part B: Fundamentals, 2001, 40(6): 473-493.
[24]	EIVAZI H, VEISI H, NADERI M H, et al. Deep neural networks for nonlinear model order reduction of unsteady flows[J]. Physics of Fluids, 2020, 32:DOI:10.1063/5.0020526.
[25]	EIVAZI H, LE-CLAINCHE S, HOYAS S, et al. Towards extraction of orthogonal and parsimonious non-linear modes from turbulent flows[J]. Expert Systems with Applications, 2022, 202:DOI:10.1016/j.eswa.2022.117038.
[26]	KINGMA D P, WELLING M. Auto-encoding variational bayes[J]. arXiv Preprint arXiv, 2013:DOI:10.48550/arXiv.1312.6114.
[27]	WU Haixu, HU Tengge, LIU Yong, et al. Timesnet: temporal 2d-variation modeling for general time series analysis[J]. arXiv Preprint arXiv, 2022:DOI:10.48550/arXiv.2210.02186.

参数	数值
泄漏孔径L/cm	10
环境风速u/(m·s^-1)	0,1,2,3,4,5,6,8,10
泄漏质量流量Q/(kg·s^-1)	5、6、7、8、9、10、15、20

参数名称	数值
输入时序长度	8
预测时序长度	16
模型层数	2
模型维数	32
周期分解数量	5
前馈网络维度	1 024