PDF下载

[1]杨阳,李成志,杜选,等.基于KNN和随机森林算法的腐蚀泄漏风险软检测模型[J].油气储运,2024,43(08):1-14.
　YANG Yang,LI Chengzhi,DU Xuan,et al.Soft detection model of corrosion leakage based on KNN and random forest algorithm[J].Oil & Gas Storage and Transportation,2024,43(08):1-14.

点击复制

基于KNN和随机森林算法的腐蚀泄漏风险软检测模型

《油气储运》[ISSN:1000-8241/CN:13-1093/TE] 卷: 43 期数: 2024年08期页码: 1-14 栏目: 出版日期: 2024-08-25

Title:: Soft detection model of corrosion leakage based on KNN and random forest algorithm

作者:: 杨阳; 李成志; 杜选; 于晓; 董绍华; 1.昆仑数智科技有限责任公司；2.华北电力大学能源动力与机械工程学院；3.中国石油大学（北京）安全与海洋工程学院

Author(s):: YANG Yang; LI Chengzhi; DU Xuan; YU Xiao; DONG Shaohua; 1. Kunlun Digital Technology Co., Ltd.;2. School of Energy Power and Mechanical Engineering, North China Electric Power University;3. School of Safety and Marine Engineering, China University of Petroleum in Beijing

关键词:: 城镇燃气管网; KNN; 随机森林; 泄漏风险; 软检测; 定量风险评价

Keywords:: city gas pipeline network; KNN; random forest; leakage risk; soft detection; quantitative risk evaluation

分类号:: TE988

文献标志码:: A

摘要:: 【目的】城镇燃气管网完整性管理需要有效的风险评价方法，腐蚀泄漏风险评价需要将风险评价因子充分与各项检测业务进行结合，然而当前检测数据繁杂且缺失严重，亟需一种可预测与评价腐蚀泄漏风险的方法。【方法】通过相关性分析，筛选出与腐蚀泄漏风险相关的关键指标，结合管道本体数据和周围环境数据，采用KNN和随机森林算法，建立智能软检测模型。【结果】该模型能够对缺失检测数据进行预测，实现关键指标的间接测量，模型预测值与真实测量值的相对误差小于25%，达到合格水平。该模型可在数据缺失情况下有效预测管道腐蚀泄漏风险，为定量评价奠定基础。与前人研究相比，模型在多因素耦合关系提取与算法选择上进行创新，提高了预测的准确性与可靠性。然而，部分异常数据表明该模型在某些条件下的预测能力有限，且模型依赖于数据完整性和准确性。设法提高检测数据数量和质量，优化关键风险指标特征提取方法，可以进一步提高模型精度。【结论】本研究丰富了燃气管道腐蚀泄漏风险预测理论，在提高管道运行安全性和可靠性方面具有实用价值，未来应着重改进数据采集和分析技术，进一步优化模型结构，提升其在不同应用场景下的适应性和准确性。

Abstract:: Integrity management of urban gas pipeline network requires effective risk evaluation methods. Corrosion and leakage risk evaluation needs to fully integrate the risk evaluation factors with various inspection operations, however, the current inspection data are complicated and missing, and a method is needed to predict and evaluate the corrosion and leakage risk. In this study, key indicators related to corrosion and leakage risk are screened out through correlation analysis, and an intelligent soft detection model is established by using KNN and Random Forest algorithms in combination with pipeline ontology data and surrounding environment data. The model is able to predict the missing detection data and achieve indirect measurement of key indicators. The study shows that the relative error between the model prediction value and the real measurement value is less than 25%, reaching the qualified level. The model effectively predicts the risk of pipeline corrosion and leakage in the case of missing data, and lays the foundation for quantitative evaluation. Compared with previous studies, this study has innovations in multi-factor coupling relationship extraction and algorithm selection, which improves the prediction accuracy and reliability. However, some of the anomalous data indicate that the predictive ability of the model is limited under certain conditions and depends on data completeness and accuracy. Improving the quantity and quality of inspection data, as well as optimising the feature extraction methods for key risk indicators, can further improve the model accuracy. This study enriches the theory of gas pipeline corrosion and leakage risk prediction, which is of practical value in improving the safety and reliability of pipeline operation. In the future, we should focus on improving the data acquisition and analysis techniques, further optimising the model structure, and enhancing its adaptability and accuracy in different application scenarios.

参考文献/References:

[1] 金龙，陈樑，王海燕，张昇.不同土壤对输气管道泄漏扩散影响模拟分析[J].石油与天然气化工，2020，49（4）：101-108. DOI：10.3969/j.issn.1007-3426.2020.04.018.
JIN L, CHEN L, WANG H Y, ZHANG S. Simulation analysis of the influence of different soil on gas pipeline leakage and diffusion[J]. Chemical Engineering of Oil and Gas, 2020, 49(4): 101-108.
[2] 高琳，曹建国.管道泄漏检测方法研究综述[J].现代制造工程，2022（2）：154-162. DOI：10.16731/j.cnki.1671-3133.2022.02.024.
GAO L, CAO J G. Review of diagnostic technique for pipe leakage[J]. Modern Manufacturing Engineering, 2022(2): 154-162.
[3] 谭震，郭新蕾，李甲振，郭永鑫，潘佳佳.基于多尺度卷积神经网络的管道泄漏检测模型研究[J].水利学报，2023，54（2）：220-231. DOI：10.13243/j.cnki.slxb.20220644.
TAN Z, GUO X L, LI J Z, GUO Y X, PAN J J. Multi-scale convolutional neural network model for pipeline leak detection[J]. Journal of Hydraulic Engineering, 2023, 54(2): 220-231.
[4] 罗宗林.城市直埋燃气管道泄漏扩散对环境影响的研究[D].哈尔滨：哈尔滨工业大学，2021.
LUO Z L. Study on environmental impact of leakage and diffusion of urban directly buried gas pipeline[D]. Harbin: Harbin Institute of Technology, 2021.
[5] HU Z K, CHEN B Q, CHEN W L, TAN D B, SHEN D T. Review of model-based and data-driven approaches for leak detection and location in water distribution systems[J]. Water Supply, 2021, 21(7): 3282-3306. DOI: 10.2166/ws.2021.101.
[6] 戴海霞，李文.基于Spearman秩相关系数与AHP法的天然气管道运行风险绩效评估[J].山东化工，2023，52（2）：178-181. DOI：10.3969/j.issn.1008-021X.2023.02.050.
DAI H X, LI W. Natural gas pipeline operation risk performance evaluation based on Spearman’s rank correlation coefficient and AHP method[J]. Shandong Chemical Industry, 2023, 52(2): 178-181.
[7] 姜礼涛，周爱红，刘超.基于灰色关联度的隧洞围岩分类研究[J].河北地质大学学报，2022，45（4）：58-64. DOI：10.13937/j.cnki.hbdzdxxb.2022.04.009.
JIANG L T, ZHOU A H, LIU C. Study on classification of tunnel surrounding rock based on grey correlation degree[J]. Journal of Hebei GEO University, 2022, 45(4): 58-64.
[8] 赵源上，林伟芳.基于皮尔逊相关系数融合密度峰值和熵权法典型场景研究[J].中国电力，2023，56（5）：193-202. DOI：10.11930/j.issn.1004-9649.202212032.
ZHAO Y S, LIN W F. Research on typical scenarios based on fusion density peak value and entropy weight method of Pearson’s correlation coefficient[J]. Electric Power, 2023, 56(5): 193-202.
[9] 纪德洋，金锋，冬雷，张姗，于坤洋.基于皮尔逊相关系数的光伏电站数据修复[J].中国电机工程学报，2022，42（4）：1514-1522. DOI：10.13334/j.0258-8013.pcsee.211172.
JI D Y, JIN F, DONG L, ZHANG S, YU K Y. Data repairing of photovoltaic power plant based on Pearson correlation coefficient[J]. Proceedings of the CSEE, 2022, 42(4): 1514-1522.
[10] SELVY P T, RAGUL M, VIGNESH G N, ANITHA M. Real time efficient accident predictor system using machine learning techniques (kNN, RF,LR, DT)[J]. International Journal of Engineering and Advanced Technology, 2020, 10(2): 108-111. DOI: 100.1/ijeat.D6910049420.
[11] 宫彦双，谷坛，安超，张爱良，廖柯熹.基于Pearson相关系数的集输管道流动腐蚀主控因素分析[J].石油与天然气化工，2021，50（2）：93-99. DOI：10.3969/j.issn.1007-3426.2021.02.016.
GONG Y S, GU T, AN C, ZHANG A L, LIAO K X. Analysis of main control factors for flow corrosion of gathering and transportation pipeline based on Pearson correlation coefficient[J]. Chemical Engineering of Oil and Gas, 2021, 50(2): 93-99.
[12] 陈世群，高伟，陈孝琪，涂彦昭，杨艳.一种基于极限学习机和皮尔逊相关系数的光伏阵列故障快速诊断方法[J].电气技术，2021，22（10）：57-64. DOI：10.3969/j.issn.1673-3800.2021.10.010.
CHEN S Q, GAO W, CHEN X Q, TU Y Z, YANG Y. A fast fault diagnosis method for photovoltaic array via extreme learning machine and Pearson’s correlation coefficient[J]. Electrical Engineering, 2021, 22(10): 57-64.
[13] 姚梦月.埋地油气管道外腐蚀速率及剩余寿命预测模型优化研究[D].西安：西安建筑科技大学，2020.
YAO M Y. Study on the optimization of prediction model for the external corrosion rate and residual life of buried oil and gas pipelines[D]. Xi’an: Xi’an University of Architecture and Technology, 2020.
[14] XUE X M, ZHOU J Z. A hybrid fault diagnosis approach based on mixed-domain state features for rotating machinery[J]. ISA Transactions, 2017, 66: 284-295. DOI: 10.1016/j.isatra.2016.10.014.
[15] ZHANG J, BIAN Z K, WANG S T. Self-paced and Bayes-decision-rule linear KNN prediction[J]. International Journal of Machine Learning and Cybernetics, 2022, 13(11): 3267-3283. DOI: 10.1007/s13042-022-01593-9.
[16] 郭小萍，滕佳岐，李元.基于EEMD-kNN的工业过程微小故障检测[J].计算机应用与软件，2023，40（4）：36-40，53. DOI：10.3969/j.issn.1000-386x.2023.04.007.
GUO X P, TENG J Q, LI Y. Incipient fault detection for industry process based on EEMD-kNN[J]. Computer Applications and Software, 2023, 40(4): 36-40, 53.
[17] 刘培栋，刘立巍，李明，赵光，刘书元，邵壮，等.基于KNN和RF结合的供热机组热负荷预测研究[J].全球能源互联网，2021，4（6）：595-601. DOI：10.19705/j.cnki.issn2096-5125.2021.06.009.
LIU P D, LIU L W, LI M, ZHAO G, LIU S Y, SHAO Z, et al. Research on heat load forecasting of heating unit based on combination of KNN and RF[J]. Journal of Global Energy Interconnection, 2021, 4(6): 595-601.
[18] 陈智鹏.基于遗传KNN聚类的机械加工故障趋势预测模型[J].机械与电子，2019，37（5）：59-62，70. DOI：10.3969/j.issn.1001-2257.2019.05.014.
CHEN Z P. Fault trend prediction model of machining based on genetic KNN clustering[J]. Machinery & Electronics, 2019, 37(5): 59-62, 70.
[19] 李洪奇，杨中国，朱丽萍，刘蔷.基于数据集特征的KNN最优K值预测方法[J].计算机应用与软件，2016，33（6）：54-58，83. DOI：10.3969/j.issn.1000-386x.2016.06.014.
LI H Q, YANG Z G, ZHU L P, LIU Q. Prediction method of optimal K value in KNN based on dataset features[J]. Computer Applications and Software, 2016, 33(6): 54-58, 83.
[20] 林艳竹，韩忠，黄林显，邢立亭，梁浩，侯金霄.基于随机森林法的弥河-潍河流域地下水质量评价研究[J].水文，2023，43（3）：60-64，70. DOI：10.19797/j.cnki.1000-0852.20210513.
LIN Y Z, HAN Z, HUANG L X, XING L T, LIANG H, HOU J X. Research on groundwater quality assessment of Mihe-Weihe River Basin based on random forest algorithm[J]. Journal of China Hydrology, 2023, 43(3): 60-64, 70.
[21] MADANI A, HAGAGE M, ELBEIH S F. Random forest and logistic regression algorithms for prediction of groundwater contamination using ammonia concentration[J]. Arabian Journal of Geosciences, 2022, 15(20): 1619. DOI: 10.1007/s12517-022-10872-2.
[22] ZHANG H Z, ZIMMERMAN J, NETTLETON D, NORDMAN D J. Random forest prediction intervals[J]. The American Statistician, 2020, 74(4): 392-406. DOI: 10.1080/00031305.2019.1585288.
[23] 骆正山，杨枚桧，王小完，张新生.基于RF-GOA-RVM的海底管道腐蚀速率预测[J].消防科学与技术，2021，40（9）：1337-1340. DOI：10.3969/j.issn.1009-0029.2021.09.016.
LUO Z S, YANG M H, WANG X W, ZHANG X S. Prediction of corrosion rate of submarine pipelines based on RF-GOA-RVM[J]. Fire Science and Technology, 2021, 40(9): 1337-1340.
[24] 陈一诺，董绍华.基于随机森林算法的管道缺陷预测方法[J].油气储运，2018，37（9）：975-979. DOI：10.6047/j.issn.1000-8241.2018.09.003.
CHEN Y N, DONG S H. Pipeline defect prediction method based on Random Forests algorithm[J]. Oil & Gas Storage and Transportation, 2018, 37(9): 975-979.

备注/Memo

基金项目：中国石油天然气集团有限公司项目“天然气管网及站场智能管控系统研发”，2021DJ7304。
作者简介：杨阳，男，1986年生，工程师，2013年硕士毕业于西南石油大学油气田材料与应用专业，现主要从事油气管道完整性管理、材料科学与工程等专业方向的研究工作。地址：北京市昌平区沙河西沙屯中石油科技园12地块B1座，102206。电话：18210920815。Email：yangyang2021@cnpc.com.cn

更新日期/Last Update: 2024-06-14