安全、健康和环境2025，Vol.25Issue(8)：27-36,10.DOI:10.3969/j.issn.1672-7932.2025.08.004
基于广义似然比的管道泄漏集成学习检测方法
An Ensemble Learning Detection Method for Pipeline Leakage Based on Generalized Likelihood Ratio
摘要
Abstract
Aiming at the problems existing in the traditional weak leakage detection methods of gas trans-mission pipelines,such as the difficulty in model construction,the scarcity of leakage data,and the excessive false alarms for minor leaks,an ensemble learning detection method for pipeline leakage based on generalized likelihood ratio was proposed.Firstly,the pipeline operation state inspection statistics based on the generalized likelihood ratio were constructed.Secondly,the pipeline inlet and outlet pressure data and outlet flow rate data were obtained,and an alarm threshold was selected.Finally,based on the initial assessment of leakage from three types of real-time monitoring data,an ensemble learning voting method was applied to integrate the results of the three leakage detection methods.The ability of the model to reduce false alarms was validated using a gas pipeline leakage test bench.The results showed that the minimum leakage detection amount of the proposed mod-el can reach 0.5%of the pipeline transportation volume,the leakage detection rate can attain 100%,and the number of false alarms can be reduced to zero.The pipeline leakage ensemble learning detection method based on generalized likelihood ratio has the advantages of simple model construction,low data requirements,and high leakage detection accuracy,providing a new idea for the detection of micro-leaks in gas transmission pipelines.关键词
集成学习/输气管道/泄漏预警/假设检验/广义似然比Key words
integrated learning/gas pipeline/leakage warning/hypothesis testing/generalized likelihood ratio分类
能源科技引用本文复制引用
刘涛,蔡秀全,周伟,王魁涛,王金江..基于广义似然比的管道泄漏集成学习检测方法[J].安全、健康和环境,2025,25(8):27-36,10.基金项目
新型油气勘探开发国家科技重大专项(2024ZD1403305),深水干式油气生产处理平台安全风险评估技术体系研究 （2024ZD1403305）
中海油十四五重大科技项目课题(KJGG-2023-17-01),基于数据挖掘分析的风险评估及泄漏预测技术研究. （KJGG-2023-17-01）
