Gas pipeline network leak localization method based on transfer learning

doi:10.16265/j.cnki.issn1003-3033.2025.03.0957

Abstract

Abstract:

In order to enhance the reliability and safety of gas network operations and improve the fault diagnosis capabilities for gas network leaks, while addressing issues such as the scarcity of real gas network leak data samples and variations in operating conditions, a gas network leak localization method based on transfer learning was proposed. Firstly, the Random Forest feature importance ranking method was used to select five pressure monitoring points in the TGNET simulation network. Subsequently, pressure monitoring point data under three different pressure conditions were respectively used as the source domain and target domain input features. The traditional JDA method of transfer learning was improved to reduce the feature distance between the source domain and the target domain. Furthermore, the CS algorithm was employed to optimize the dimensionality after mapping d' and the learning rate λ parameters of the improved transfer learning algorithm, ultimately achieving the diagnosis of unlabeled target domain leak segments. The results indicated that the proposed leak localization method for complex gas networks can effectively improve the localization accuracy of unlabeled gas network leaks, achieving higher accuracy compared to traditional.

Key words: transfer learning, gas pipeline network, leak localization, random forest, cuckoo search(CS) algorithm, joint distribution adaptation(JDA)

CLC Number:

X928.03事故预防与预测

CHEN Cen, JI Yubo, WANG Huan, NIE Rongshan, LIANG Xiaoyu. Gas pipeline network leak localization method based on transfer learning[J]. China Safety Science Journal, 2025, 35(3): 212-220.

Figures/Tables 12

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Table 1

Table 2

Table 3

Fig.7

Table 4

Fig.8

References 19

[1]	LIU Yiming, BAO Yi. Review on automated condition assessment of pipelines with machine learning[J]. Advanced Engineering Informatics, 2022,53:1-5.
[2]	向鹏成, 张子薇, 李卉, 等. 城市地下燃气管道泄漏事故致因因素系统分析[J]. 中国安全科学学报, 2023, 33(12):140-147. doi: 10.16265/j.cnki.issn1003-3033.2023.12.2162
	XIANG Pengcheng, ZHANG Ziwei, LI Hui, et al. Systematicanalysis on causative factors of gas pipelines leakage accidents[J]. China Safety Science Journal, 2023, 33(12): 140-147. doi: 10.16265/j.cnki.issn1003-3033.2023.12.2162
[3]	KORLAPATI N V S, KHAN F, NOOR Q, et al. Review and analysis of pipeline leak detection methods[J]. Journal of Pipeline Science and Engineering, 2022, 2(4): 1-17.
[4]	康俊鹏. 基于压力信号处理的油气管道泄漏检测技术[J]. 石化技术, 2015, 22(11):93.
	KANG Junpeng. Oil and gas pipelines leakage detection based on pressure signal processing[J]. Petrochemical Industry Technology, 2015, 22(11): 93.
[5]	GUOTA P, GOYAL A, DAUWELS J, et al. Bayesian detection of leaks in gas distribution networks[C]. IECON 42^nd Annual Conference of the IEEE Industrial Electronics Society, 2016: 855-860.
[6]	贾文龙, 孙溢彬, 汤丁, 等. 基于支持向量机的输气管道泄漏压降信号智能识别方法[J]. 化工进展, 2022, 41(9):4713-4722. doi: 10.16085/j.issn.1000-6613.2021-2326
	JIA Wenlong, SUN Yibin, TANG Ding, et al. Intelligent identification method of pressure drop signals for gas pipeline leakage based on support vector machine[J]. Chemical Industry Progress, 2022, 41(9): 4713-4722.
[7]	YANG Tao, YU Xia, MA Ning, et al. Deep representation-based transfer learning for deep neural networks[J]. Knowledge-Based Systems, 2022,253:1-6.
[8]	朱良玉, 陶林, 胡超凡, 等. 基于IFOA-TCA的跨工况轴承故障诊断[J]. 轴承, 2023(4):73-79.
	ZHU Liangyu, TAO Lin, HU Chaofan, et al. Fault diagnosis for bearings under crossed operating conditions based on IFOA-TCA[J]. Bearing, 2023(4): 73-79.
[9]	JIAO Jinyang, ZHAO Ming, LIN Jing, et al. Residual joint adaptation adversarial network for intelligent transfer fault diagnosis[J]. Mechanical Systems and Signal Processing, 2020,145:1-4.
[10]	郭慧, 刘明艳. 基于遗传算法和随机森林的入侵检测方法研究[J]. 计算机应用与软件, 2024, 41(1):304-309,314.
	GUO Hui, LIU Mingyan. Intrusion detection method based on genetic algorithm and random forest[J]. Computer Applications and Software, 2024, 41(1): 304-309,314.
[11]	李红霞, 徐浩冉, 田水承. 基于随机森林的矿工不安全行为预测预警模型[J]. 中国安全科学学报, 2022, 32(12):10-18. doi: 10.16265/j.cnki.issn1003-3033.2022.12.2752
	LI Hongxia, XU Haoran, TIAN Shuicheng. A prediction and early warning model of miners unsafe behavier based on random forest[J]. China Safety Science Journal, 2022, 32(12): 10-18. doi: 10.16265/j.cnki.issn1003-3033.2022.12.2752
[12]	姜宇迪. 基于无监督深度迁移学习的电梯制动器实时状态监测和预警功能安全系统研究[D]. 上海: 上海交通大学, 2020.
	JIANG Yudi. Functional safety system of elevator brake real-time status monitoring and prognostics based on unsupervised deep transfer learning[D]. Shanghai: Shanghai Jiaotong University, 2020.
[13]	ANVAR A A T, MOHAMMADI H. A novel application of deep transfer learning with audio pre-trained models in pump audio fault detection[J]. Computers in Industry, 2023,147:1-5.
[14]	WANG Pengju, LU Luxi, LI Jian, et al. Transfer learning with joint distribution adaptation and maximum margin criterion[J]. Journal of Physics: Conference Series, 2019, 1169(1):1-9.
[15]	LI Yibai, WANG Haoyun, ZHANG Yuzhuo, et al. Inversion of the optical properties of apples based on the convolutional neural network and transfer learning methods[J]. Applied Engineering in Agriculture, 2022, 38(6): 931-939.
[16]	XING Bo, GAO Wenjing, MARWLA T. An overview of cuckoo-inspired intelligent algorithms and their applications[C]. 2013 IEEE Symposium on Swarm Intelligence, 2013: 85-89.
[17]	WANG Lijin, ZHONG Yiwen. Cuckoo search algorithm with chaotic maps[J]. Mathematical Problems in Engineering, 2015, 2015(1):1-14.
[18]	李兴泉, 马玉兴. 基于TGNET软件的天然气管网动静态模拟分析[J]. 节能, 2023, 42(4):19-21.
	LI Xingquan, MA Yuxing. Dynamic and static simulation analysis of natural gas pipeline networks based on TGNET software[J]. Energy Conservation, 2023, 42(4): 19-21.
[19]	LAURENS V D M, HINTON G. Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008, 9:2579-2605.

气源压力/MPa	泄漏标签	样本数
0.4	1-16	2 391
0.3	1-16	2 391
0.2	1-16	2 391

参数	下限	上限	最优解
d'	10	100	16
λ	0.001	1	0.597

源域工况压力/MPa	目标域工况压力/MPa	准确率/%	平均准确率/%
0.4	0.3	99.85	97.79
0.4	0.2	95.73	97.79
0.3	0.2	99.97	99.97
0.3	0.4	99.96	99.97
0.2	0.3	99.92	93.08
0.2	0.4	86.24	93.08

任务	SVM	随机森林	K最近邻	文中方法
0.4 MPa→0.3 MPa	36.96	37.68	35.98	99.85
0.4 MPa→0.2 MPa	36.96	37.68	35.98	95.73
0.3 MPa→0.2 MPa	36.99	37.29	36.07	99.97
0.3 MPa→0.4 MPa	36.99	37.29	36.07	99.96
0.2 MPa→0.3 MPa	37.02	37.57	36.43	99.92
0.2 MPa→0.4 MPa	37.02	37.57	36.43	86.24