基于流体力学的燃气管网泄露检测研究

作     者：许鑫 郑博豪 李星阳 

作者机构：泰州市市域社会治理现代化指挥中心江苏 泰州 浙江理工大学计算机科学与技术学院浙江 杭州 

出 版 物：《人工智能与机器人研究》 (Artificial Intelligence and Robotics Research)

年 卷 期：2025年第14卷第1期

页      面：183-193页

学科分类：08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：燃气管网 流体力学 PIPENET 泄露检测 反向估计算法 

摘      要：本文提出一种快速准确的新型燃气管道泄漏检测方法。该方法利用PIPENET进行管道建模,引入基于流体力学原理的反向估计算法,根据管段瞬时压力、流量数据,形成瞬态测量数据进行泄漏定位、诊断、状态估计等在线诊断技术。本研究测试了模拟数据与某工业园区下游15个用户节点的真实数据,泄露正确检出率平均在90%以上,这一实验结果表明本研究的检测技术有较高的可适用性,该研究可为城镇燃气管网泄露检测提供技术支持。This paper proposes a fast and accurate new method for detecting gas pipeline leaks. This method utilizes PIPENET for pipeline modeling and introduces a reverse estimation algorithm based on fluid mechanics principles. Based on the instantaneous pressure and flow data of the pipeline section, transient measurement data is formed for online diagnostic techniques such as leak location, diagnosis, and state estimation. This study tested simulated data and real data from 15 downstream user nodes in an industrial park. The average correct detection rate of leaks was over 90%. This experimental result shows that the detection technology in this study has high applicability and can provide technical support for leak detection in urban gas pipelines.