第38卷第4期 2017年4月
仪器仪表学报 Chinese Jourmal of Scientific Instrument
Vol. 38 No. 4 Apr.2017
基于Markov特征的油气管道泄漏检测与定位方法
刘金海，藏东，汪刚
（东北大学信息科学与工程学院沈阳110004）
摘要：针对传统的基于压力信号的管道泄漏检测方法误报率和漏报率偏高，同时定位误差较大的缺点，设计了一种基于 Markov特征的管道泄漏检测与定位方法。首先，将管道压力数据构造为Markov链的形式，并提取其动态特征；然后，将所提取的特征应用于Neyman-Pearson异常检测方法之中，检测全部压力数据样本的状态，并对检测到的异常样本进行同源信号匹配，修正检测结果；最后，将相似性定位方法与连续小波定位方法结合，确定管道首末两端响应压力变化的时间差，并根据管道长度和压力波传输速度等信息，对泄漏源定位。所提方法能应用于小泄漏和缓慢泄漏的检测与定位，易于实现，误报率与漏报率显
著降低，定位精度提高。通过对历史数据的分析，验证了所提方法的可行性和有效性。关键词：油气管道；Markov链；Neyman-Pearson异常检测；相似性定位；连续小波定位
中图分类号：TH878TH865TE832TE88
文献标识码：A
国家标准学科分类代码：440.55
Leakage detection and location method of oil and gas pipelines
based on Markov features Liu Jinhai, Zang Dong, Wang Gang
(School of Information Science and Engineering, Northeastern University, Shenyang I10004, China)
Abstract : Aiming at shortcomings of traditional leakage detection methods based on pressure signal, such as high false positive rate and false negative rate, and large location error, a leakage detection and location method of oil and gas pipeline based on Markov feature is designed. First of all, a Markov chain is built from the pipeline pressure data, at the same time, the dynamic features are extracted; then, in order to detect all pressure data sample states, the Neyman-Pearson anomaly detection method is applied to the dynamic features; meanwhile, the detection results are corrected through checking whether the detected abnormal samples are matching with the same source signal; last, similarity location method and continuous wavelet location method are combined to determine the time difference of the pressure changes at both ends of the pipe, and the leakage is located according the information of pipeline length and pressure wave transmission speed. The method presented in this paper can be used in the detection and localization of little leakage and slow leakage, is easy to implement; moreover, the false positive rate and false negative rate are significantly decreased, the positioning accuracy is improved. The historical data analysis verifies the effectiveness and feasibility of the proposed method.
Keywords : oil and gas pipeline ; Markov chain; Neyman-Pearson anomaly detection ;similarity location ; continuous wavelet location
气管道故障的有效检测和及时处理已迫在眉睫。
1引言
目前世界上50%的油气管道已经使用了30年甚至更长时间，由于腐蚀、第三方的破坏等原因，管道泄漏事故时有发生。因为油气管道所输送介质的危险性和污染性，一且发生泄漏事故，轻则造成能源原料的流失，影响经济效益，污染环境，重则导致人身伤害事故，因此对油
在众多的泄漏检测方法中，可大致分为管道外部环
境检测、管壁状况检测以及管道内部流体状态检测3 类。基于光纤[2]、声波3]等方法的泄漏检测系统虽然已经取得一定应用效果，但是基于压力波方法的泄漏诊断系统因其安装简单、信号获取方便、成本低、精度高、维护容易、适用广泛等优点，依然被该领域的工程人员大规模使用。具不完全统计在我国至少有两万公里以上的长
收稿目期；2016-08
Received Date: 2016-08
*基金项有：有数据科学基金（61473069，61627809,61374124）、中央高校基本科研业务费专项资金（N140405002）项目资助